G0Analysis Release 0.01
     Jason Lenoble; 2001/11/26 20:31:23.

     Initial CVS release of the G0 Analysis softwre.

G0Analysis Release 0.02
     Paul King; 2002/02/11 17:39:38.

     Corrected several minor compilation problems so that G0Analysis
     now compiles on the JLab Linux machines (ifarml1, etc.) and on
     the JLab Solaris machines (jlabs2, ifarms1, etc.).

G0Analysis Release 0.03
     Paul King; 2002/02/18 17:39:00.

     Added ".keepme" files to the directories: "Analysis/dictionary",
     "G0ROOT/dictionary", "G0ROOT/scripts", "GUI/dictionary", "bin",
     and "lib".  These files will ensure that the directory is created
     when the source code is checked out using the "-r" or "-P" options.

     Also, this release history was implemented.

     Other changes since release 0.02:
     * spayde on Fri Feb 15 11:40:10 EST 2002
       Modified Files:
		.G0Setup.bash .G0Setup.csh
     * pking on Mon Feb 18 15:51:26 EST 2002
       Added administrative files to control loggin of the taggin utilities.

G0Analysis Release 0.04
     Paul King; 2002/03/05 11:45.
     
     This will be the last release before switching to the "new" GNUMakefile
     prepared by Jason.

     Other changes since release 0.03:
     * Modified the setup scripts so that if the G0SCRATCH directory does
       not exist, an error is relayed to the user.
       Added a setup script for CODA which runs under the bash shell.
     * Modifed the GetPhysicsEvent to process ROC2 and ROC4 as North
       American scaler crates.
     * Compiles the code against libz.so.
     * Updated the MySQL database interface to reflect the lastest
       database schema (v0.5).
     * Major change in how the NA scalers are decoded.
     * Conditionally includes the ntuples for commisioning.
     * Modified TAnalysis so that the French ntuple is not included in
       the G0ROOT G0Analysis Tree if there are no entries in the ntuple.
     * Several changes in "#include" usage in the modules; this was in
       preparation for the new "modular" GNUMakefile.
     * Added fuctionality to fill analysis table in G0 database (G0
       database schema v. 0.5).
     * Transfered path handling and root file handling from class TPilot
       to new metaclasses MG0PathHandler and MG0ROOTFileHandler.
     * Added a definition for "gzerol3" to the ET client routines.
     * Precompiler flags were added to allow conditional inclusion of
       the Fastbus ntuple.

G0Analysis Release 0.05
     Paul King; 2002/03/13 19:23.
     
     This release is the first version which contains the GNUmakefile
     as produced by Jason.
     In a later release, the old "Makefile*" files will be removed.

     Additional changes:
     * Added the classes CalcRates, CalcYieldAve, and UpdateRateFile
       and the GUI main routine G0RateMonitor to allow a 'visual
       scaler' type display for the events within specified time
       windows.
       Several other files were modified to support this functionality.
     * The database routines were refined further.

G0Analysis Release 0.06
     Paul King; 2002/03/28 11:20.

     This release has several modifications.

     Changes:
     * Modified the beam and MPS ntuples.
       Added a fake cut on the beam current in "PassedCuts.C".
       Minor correction in G0Analysis main loop;  it had been
       counting the number of event one more that the actual. 
     * Correction to GNUmakefile; variable DEBUGS was expected
       to be named DEBUG within the makefile, so renamed it to
       have the debug flag taken into account
     * Linking bug when g0root is built with flag -D__NOFASTBUS
       fixed by including file TSetupProfiles.h to help the
       makefile during the config step
     * Revised decoding for French ntuple (indexes mismatch)
     * Modifed the fG0FileInfo member of the MG0ROOTFileHandler
       class; changed MG0FileInfo, can use "PrintInfo" to see
       all of the info about analysis parameters and run setup as
       passed from g0analysis through MG0FileInfo to g0root.
       Extended the calss to pass more information about the
       analysis parameters.
     * Changed GetPhysicsEvent to catch the French configuration
       event.
     * Several changes to g0root...
     * Fixed backward incompatability introduced with new version
       of class MG0FileInfo : old ROOT files couldn't be read anymore.
       Fixed also a problem when a ROOT file is transfered as is from
       one computer to another: then the setup file path under
       $G0SCRATCH is likely to be inaccurate; now before aborting,
       g0root tries to find the file also in the default setup directory
       under $G0ANALYSIS
     * Put in a normalization routine for the French yields.
       Set up the beam scaler map file decoding.
       Added an upper limit to the number of entries in the commisioning
       ntuples; the limit is defined in G0Globals.
       Updated G0BeamMonitor to remove redundant variables.
     * Added protection  for NA and beam bank reading when one or more
       scalers are dead.
     * Revised the asymmetry and average yield calculation. Fixed several
       bugs in the G0French assignment routine and the normalizing yield
       routine. Yield & asymmetry for Beam, NA and French class and their
       running averages and sigma calculations were checked by hand.
       Fixed several bugs in the ntupling routine.
       Added the 1st bin (counts accumulated from -6.5ns to 0 ns) for NA
       data into the TOF histograms.
       Made the running average and the sigma calculation more
       object-orientated so that they only take Helicity or Asymmetry
       objects.  Also defined the overloaded running averaging operators
       in Beam,NA and French class. This proved to improve the speed of
       the code.
     * Minor change in "TSetupColumn.h" to use the standard "#include"
       notation for the iostream.h
     * Added an installation file with a discussion of the usage of the
       GNUmakefile.

G0Analysis Release 0.07
     Paul King; 2002/05/21 11:30.

     Changes:
     * Updated the NA scaler map file to include all three crates.
     * Enabled the code to ignore a new Fastbus decoding word.
     * Various changes to clean up the code:
       1) Did the speed test on the offline code with Wolfgang. Optimized
	  the code at various places. Modified NA ProcessEvent so that it
	  no longer does the dividing of the time bin width (especially
	  the 0th timebin with 6.5 ns width). Dividing the yield at this
	  step creates round-off error.
       2) Cleaned up vdetector/tscintillator/octant routines. Provided
	  the "cleaners" for the root objects in NA, French, Beam,
	  Asymmetry and Helicity classes to avoid potential segv.
       3) Studied the NA deadtime correction a little more. Implement
	  Wolfgang's changes. Also changed the convergence limit to %5
	  of statistical error, based on a simulation, to reduce the
	  number of the required iterations.
       4) Set the flag to turn off the histogramming for NA, French,
	  Beam, Asymmetry and Helicity classes. The command line flag:
	  --offhists BeNaFrAsYi. (The order of the objects is not
	  important.)
	  Also create TDirectories for these histograms according to
	  the --offhists flag and sort the histograms into them. The
	  dirs are: "NATOF", "FrenchTOF", "BeamHists", "AsymHists",
	  "YieldHists".
       5) Multi-run replay (-r firstrun:lastrun) fixed and tested.
	  Fixed some loop counters and misc flags for the purpose of
	  the multi-run replay.
       6) Now no error messages for the size of TS crate when doing\
	  fastbus run replay. 
     * Entered several new Fastbus channel mapping files.
       The current list of setup files is described in this table.
       ------------------------------------------------------------
       Beginning run	Ending run	Configuration File
	  6728		  7311		SetupChainTest.05-25-01.dat
	  7312		  8808		SetupChainTest.06-25-01.dat
	  8809		  8869		Setup.11-07-01.dat
	  8870		  9146		Setup.12-04-01.dat
	  9147		  9945		Setup.01-15-02.dat
	  9946		  9950		Setup.04-12-02.dat
	  9951		  9956		Setup.04-16-02.dat
	  9957		  9960		Setup.04-17-02_early.dat
	  9961		  9987		Setup.04-17-02.dat
	  9988		  9995		Setup.04-30-02.dat
	  9996		999999		Setup.05-06-02.dat
       ------------------------------------------------------------
     * Changed the name of the real-time executable from "g0online" to
       "g0realtime".
     * Consolidated all of the file opening operations into a new
       function, to aid readability of the source code.
     * Created a function ExitGracefully. This function closes files,
       etc. in case of problems.
     * The main G0Analysis routine and the main G0Realtime routine have
       been modified to use the Fastbus configuration file "Setup.05-06-02.dat"
       by default if the -s flag is not used on the command line.
     * More code clean-up:
       1) Removed the chain test and system test remainders in the code
	  to avoid  confusion.
       2) Removed the flag to switch off TS data analysis. We will always
	  analyze the TS crate data.
       3) Moved the yield normalization against the beam charge after the
	  event cut on the beam charge. Also reset some quartet and event
	  counters before the run-by-run analysis for proper multi-run replay.
       4) Removed ConstructHists(), FillHists() and ClearHists() methods
	  in NA, French, Beam, Yield, Asym classes. Defined a new G0Hists
	  class that handles the histogramming of all the physical classes,
	  except for Fastbus.
       5) By default, no histogramming during the on/offline replay.
	  Added --onhists <ring#> flag on the command line to turn it on,
	  if desired.
	  The histograms are defined in the G0Hists class according to the
	  list Paul sent out (FastBus hists are left out).
       6) Added a Sigma function (as inline function.h in sample code)
	  routine to to make the calculation of standard deviation in
	  different classes a little more compact. 


G0Analysis Release 0.08
     Paul King; 2002/05/21 11:30.

     Version frozen just before addition of the new Fastbus routines.
     Changes:
     *  Added the error scalers to the "nascalerMap.in" file and updated
	the NA detector class and the decoding routines to upack them from
	the data stream.

G0Analysis Release 0.09
     Paul King; 2002/07/18 14:20.

     Version frozen just before several obsolete files are removed, and
     the GNUmakefile is modified to not compile the g0root and g0monitor
     executables.
     The obsolete files to be removed are:
	   Analysis/src/Asymmetries.C
	                CompAsyms.C
	                CutOnYield.C
	                NSigmaDiffCut.C
	                SortPosNeg.C
	                TFrenchEvent.C
	   Analysis/include/Asymmetries.h
	                    CompAsyms.h
	                    CutOnYield.h
	                    NSigmaDiffCut.h
	                    SortPosNeg.h
	                    TFrenchEvent.h

     Changes:
     *  The French detector TOF decoding was rewritten by Dominique and
	Jianglai.
     *  The hooks for G0ROOT were removed from the Analysis directory.
	The Fastbus class was rewritten; the decoding of fastbus buffer
	is done in a similar way as NA and Beam buffers. Took the original
	setupfiles of G0ROOT as the fastbus channel maps.
     *  Added the Fastbus ntuples to the G0Ntuple class.
     *  Redefined the ltd_error in TCommunicator as UInt_t for
	consistency.
     *  Modifed GetPhysicsEvent.C. To optimize the speed, TmpBuffers are
	no longer filled.
     *  Added some  map files in the parameter input directory.
	"nascalersMap.in.old" is the old NA scaler map for Dec. 01 test.
	"Setup.05-06-02.dat" and "Setup.12-04-01.dat" are two fastbus
	map files.
	"fastbusMap.in" now points to "Setup.05-06-02.dat".
     *  Added an input file in prminput dir dmchs.in which reflects the
	empty dmchs and missing dmchs.  Added the corresponding routine
	in French and G0ReplayPrms classes to do the reading and
	processing this file. This change reflected the hardware
	configuartion change since Dominique's last visit.
     *  Separated out a StringManip.{C,h} routines which is a bunch of
	string manipulation functions that are used when we read the
	input files.
     *  Modified G0CodaEvent to correctly skip over the "NULL" ROC buffers.
	These buffers occur when a particular ROC does not have any data to
	be read out for the current event type.
     *  Updated the fastbus configuation accordingly to include the 10 new
	channels for the French anode current and PMT base low voltage
	level monitoring. Added a "Setup.06-21-2002.dat" to reflect the 
	change. Added a frenchmonitor struct in the fastbus class.
	Modified the fastbus map file decoding for these 10 channels.
     *  In the fastbus ntuple, also added a frenchmonitor branch to hold
	the values for these 10 adc channels. Cross-checked the results
	with Christophe's hbooks and they agreed. Regroup helicity,
	gms_mask, event_type, micropulse_counter, macropulse_counter
	into a branch called "misc". 
     *  Modified G0Hists to add 512 ADC histograms into the real time mapfile.
	Not tested yet.
     *  Minor changes in the main and G0CodaEvent routines for more
	event-type sanity checks before filling the ntuples. 
     *  In CalcAsyms.C, increased the upper bound of nplus[] and nminus[]
	from 2 to 4 to prevent a potential bounds overflow during the
	helicity pattern check.
     *  Moved compiler optmization flag to the architecture dependent
	section of GNUmakefile.
     *  In G0ReplayPrms.C, replaced pow(2, 31) with (1 << 31) to get rid
	of compiler error
     *  In evio.c, typecasted a pointer and commented out a non-reachable
	statement to get rid of compiler warnings
     *  In TAnalysis.C,  Replaced (now private) method
	TStopwatch::GetRealTime()  with TStopwatch::RealTime();
	TStopwatch::Continue()
     *  Added lines in the FillFastbusNtuple method to fill the meantimer
	records in the Fastbus ntuples.  The internal variables containing
	the meantimer data were already existing, just not being used to
	fill the ntuple entries.


G0Analysis Release 0.10
     Paul King; 2002/08/23 18:45.


     Release G0Analysis 0.10 was origianlly created on 23 August 2002 at
     about 18:45.
     However, I forgot to generate the change report at that time.
     Here is the list of changes which went into release 0.10.

     Changes since release 0.09:
     ===========================
     *  Changed the default compilation targets to be only "g0analysis" and
	"g0realtime".
     *  Removed several obsolete files from the Analysis/include and
	Analysis/src directories.  These removed files are:
        Analysis/include/Asymmetries.h
                         CompAsyms.h
                         CutOnYield.h
                         NSigmaDiffCut.h
                         SortPosNeg.h
                         TFrenchEvent.h
	Analysis/src/Asymmetries.C
                     CompAsyms.C
                     CutOnYield.C
                     NSigmaDiffCut.C
                     SortPosNeg.C
                     TFrenchEvent.C
     *  Updated helicity predictor routine in G0BeamMonitor according to 
	the new algorithm of the random bit generator of G0 helicity box. 
     *  Implemented a test version of G0 SQL query translator in G0SQL
	class. Added some more string manipulation functions in
	StringManip.C. Added a G0 database lookup file G0DB_lookup.dat
	in Analysis/prminput for looking up the variable names, table
	alias and relations between tables.
     *  Implemented G0SQLResult class to handle the return values from a 
	SQL query.
     *  Reimplemented "-s" command-line flags to choose fastbus map file by 
	hand.
     *  Fixed some small bugs in G0Ntuple.C, G0Hists.C, G0NADetector.C and 
	G0FrenchDetector.C.
     *  Implemented the process by which the analyzer will automatically
	loop through the run segments when processing a segmented run.
	The segmented files will be processed in order.  The segements will
	probably still be found even if they are not consecutive.
	No error checking is done on the event number yet (this would be to 
	be sure that no segments were skipped).
     *  Updated fastbusMap.in to include the setup file "Setup.06-21-02.dat".
	Corrected both "Setup.05-06-02.dat" and "Setup.06-21-02.dat":  the 
	mapping of the French Fastbus channels had used an older scheme 
	(from the chain test) rather than the scheme with the buddy 
	detector signals interleaved.  Also, the meantimer cables were 
	installed in reverse order as compared to the initial signal layout.
     *  Added aliases for new program g0acm (G0 Anode Current Monitor)
     *  Changes in the GNUMakefile:
	- Added g0acm as default for compilation when variable CODA is
	  defined.
	- Replaced keyword dummy by xxxdummyxxx to prevent weird errors
	  when the set of environment variables is incompletely overridden.
        - Cleaning now checks for existence of directory SunWS_cache on SunOS.
	- $(ROOTCINT) is now called with flag -DROOTCINTFIX; that helps hide
	  from the ROOT script some old c standard include files referred to,
	  for example in et.h.
     *  Moved up declaration of duplicated external variables to G0Globals.h 
	to have only one point of intervention
     *  Added a warning comment for the structure FrenchMonitor, because class
	G0Hists and TGAnodeCurrentMonitor explicitly access its array members
	and would be affected by any change in size.
     *  Added several global constants for the anode current monitor
     *  In order to get compiler warnings when we try to modify them, moved
	all global variables that are not currently changed within the source
	into constants
     *  Added a few methods to encapsulate Fastbus histogram handling;
	preexisting realtime methods still process everything, but simply
	call the new methods, when it comes to Fastbus; gained a factor two
	in speed for the anode current monitor.
     *  vector<TH1F*> h_pmt_adc_ref was added as the anode current code
	needs two sets of histograms
     *  Added a new parameter file to keep the anode current calibration
	parameters; added the corresponding members and methods with respect 
	with the current model for handling parameter files
	(this_pedestalsdir,this_pedestalsfile in class G0ReplayPrms)
     *  Edited GetPhysicsEvent to have a system check when the code has
	failed to recieve data from ET for a few second : this prevents the
	anode current monitor GUI to go blank and keeps its buttons atcive
	when nothing is happening.
     *  A small bug fix preventing a tricky initialization of the G0Scint
	objects (there are twice as many of French scints as NA) from
	propagating to the entire code : after passing through the
	constructor where it is altered, kScints is now simply reset to
	the value it had when entering.
     *  10 new histograms were added corresponding to the 10 French Fastbus
	ADC monitoring channels. They are the last ten members of the
	vector h_pmt_adc, so that looping on the number of PMTs leaves them
	by default out of the loop.
     *  TGAnodeCurrentWindow.h contains a handful of classes for the anode
	current monitor program. The latter is designed as a stand-alone 
	program. It uses the common structure under Analysis/ but its 
	classes are unknown to the main code. Therefore only slight 
	adaptations were made to the main source to fit the new program.
	All classes are derived from native root objects. This was to allow
	to override methods handling mouse commands, so that the pads for
	the anode current have special behavior when clicked.
     *  Added a new readout mode for the North American decoding routines:
	"InvertedBuddies".  In this new mode, the buddy scaler channels are
	assigned to detector channels in reverse order.
	This is neeed to decode the data coming from the new SIS3801 buddy
	scalers.
	Also, corrected two decoding errors in the "Buddies" decoding
	routine, both related to the order in which the octants and
	detectors are looped over.
	The order of buddy signals for the first LTD (as an example) is:
	DT1 ==  Detector 5-01 BuddyPre,
		Number of hits in 5-01 when its buddy was busy from the
		previous micropulse;
	DT2 ==	Detector 5-01 BuddySame,
		Number of hits in 5-01 when its buddy was busy from the
		same micropulse;
	DT3 ==	Detector 1-01 BuddyPre,
		Number of hits in 1-01 when its buddy was busy from the
		previous micropulse;
	DT4 ==	Detector 1-01 BuddySame,
		Number of hits in 1-01 when its buddy was busy from the
		same micropulse.
     *  New class G0MapFile encapsulates ROOT native class TMapFile. When 
	instantiated, it checks for consistent global memory address and 
	aborts if necessary while saving to disc new higher address to be 
	used from then on by realtime executables. When it aborts, the 
	program also kills all executables previously using the erroneous 
	mapfile. They'll have to be relaunched, but they'll use as a default 
	the higher address saved to disc.
	The class does not derive from TMapFile, as the latter has no public 
	ctors. Instead it possesses a TMapFile* member, and a few methods 
	imitate the TMapFile behavior for transparency within the main 
	source. As not all methods in TMapFile have their equivalent in 
	G0MapFile, the getter GetMapFile is provided for direct manipulation 
	on the map file.
	The class G0ReplayPrms was adapted to have now a G0MapFile* member 
	rather than a TMapFile*. All codes connecting to the mapfile should 
	have now this member instead of a mere TMapFile.
	G0MapFile is currently the only class precompiled in expurged g0root.
	This is to allow to run consumer scripts while GUI codes are still
	in the process of being (re)written. The producer is always g0realtime.
	Map file size will have to be tuned. Currently it is maximized at
	10M to be certain all objects will fit in it.
     *  Changed the location of the ".MapFileBaseAddress" file from the
	environment variable $SETUP to the variable $G0_TMP_DIR.  This 
	should allow multiple machines to use the same executable version 
	while maintaining unique memory map locations.
     *  The Fastbus channel mapping files Setup.05-06-02.dat and
	Setup.06-21-02.dat have been edited by Dominique.  These changes 
	should correct the errors in the French CFD channel assignments.
     *  Reimplemented ToF histogramming for each detector (8*16). Added 
	command line flag "--tof".
     *  Merged beamPrms.in (BPMs & BCMs pedestals file) and beamscalersMap.in 
	into beam.in as a single input file.
     *  Split PassedCuts() into PassedEventCuts() to cut bad events and 
	PassedQRTCuts() to take out the bad quartet from asymmetry calculation.
     *  Added a variable NUM_CAVITY in G0Globals.C. If the configuration
	of the beam scalers is  standarized, one should change the value
	of NUM_BPM, NUM_BCM and NUM_CAVITY to reflect the real number of
	these monitors.
     *  Made gzerol3 a default producer for the anode current monitor.
	Typing command g0acm omitting flag '-h' is now OK.
     *  Minor change in the fastbusMap.in file to correct the range of
	runs for which we should use the setup file "Setup.06-21-02.dat".
	The run number has now been reset at 11000; all runs from this point
	should use the setup file "Setup.06-21-02.dat" (unless we change
	something else later on, of course).
     *  Added option "hardware_prompt" for "--H" flag to treat all 
	the reported helicity bits as prompt. When the analyzer is in
	this mode, the helicity predictor is not called.
     *  Rearranged the time-encoding event cuts into two levels. The first
	level is a rudimentary cuts on the event-type (we want 30 Hz mode)
	and the trigger quality (we expect the correct size from TS crate).
	If this cut is passed, we fill ToF, MPS and beam ntuples. The 2nd
	level will set thecut on the beam quality (charge), quartet phase,
	etc. The events passed this cut will go into the quartet analysis
	(deadtime correction, normalization, forming quartet structure,
	etc ...). 
     *  In G0Global, define global constants MAX_NUM_BPM, MAX_NUM_BCM and 
	MAX_NUM_CAVITY to give upper bounds of the number of these monitors. 
	As an ugly fix to get the numbers of different types of monitors
	from the input file, redefined the original NUM_BPM, NUM_BCM,
	NUM_CAVITY as non-constant global variables. Their value will be
	reassigned in GetBeamPrms() routine when reading the beam input
	file and be used in all the subsequent calculation and ntupling
	routines.
     *  Modified evOpen in evio.c after consultation with Steve Wood.
	1)  Changed a "case NULL" to "case '\0'";
	2)  Relocated a "{" from after the "casr 'R':" to just after the'
	    "switch (*flags)";
	3)  Removed a the trailing " BIT64" from two lines "#endif BIT64".
     *  Updated realtime histogram lists in G0Hists. We no longer build
	512 ADC histograms, since they are monitored in the stand-alone
	g0acm.
     *  Made new fastbus setupfile, Setup.08-20-02.dat. Added a channel 
	ADC10ch80 as GMSRefPMT.  Also modified relevent mapping and
	retrieving  routines. The value of this ADC channel is passed
	into G0Hists realtime histogramming routine.
     *  Made default ET host as "gzerol3" in G0RealTime.C.
     *  Implemented a EPICS event processor in G0CodaEvent class. Added
	a methods G0CodaEvent::ProcessEPICSBuffers(), in parallel with 
	G0CodaEvent::DecodeROCs(), which dumps the data buffers of the 
	epics event (maybe more than one bank) into a big string variable.
     *  Paul designed a new G0EPICSEvent class.  Implemented a
	ExtractBPMAutogains() method to get the autogain values from the
	string variable filled in ProcessEPICSBuffers(). The empty method
	ReportAutogains() meant to take the autogain values into
	G0BeamMonitor class for the position  calculations. But the string
	labels in the EPICS buffer has some discrepancy with the Junho's
	BPM map. So I will just leave it for now. The bottom line is that
	the autogain is decoded and stored in a struct in G0EPICSEvent object.
     *  G0ControlEvent class is Paul's prototype class to process the control
	events. A dummy class for now.
     *  Renamed the realtime histograms for the two striplines on G0 girder
	to the right names (G0 & G0B).
     *  Added simple-minded IsNumber() methods in StringManip.C. This method
	is currently used in the EPICS decoder. (In the data stream, if the
	server timeout happens, the field followed by the varaible name
	would be "+-", to which we do not want to assign the autogains.)
     *  New bpm calibration. 
	Pedestals with bpm gain 2000-1500 to use for beam ON data
	until we get autogains into the datastream and do the gain calibration.



G0Analysis Release 0.11
     Paul King; 2002/11/25, 14:50.

     This is the fisrt release of the G0Analysis software since the G0
     commisioning run began.
     It includes several changes which were developed during the beam
     tests as well as some fixes found during the commisioning run.

     The upcoming changes will be an improved EPICS decoding system
     and the decoding and analysis of the 120 Hz oversampling data.

     Changes since release 0.10:
     ===========================
     *  Added constants kRescaleFactor and kADCGateWidth for g0acm
	to the G0Globals.  Added text entry windows to g0acm to
	input the rescaling parmeters.  The default screen shown only
	the NA octants.
     *  Changed some names within the beam monitor class for the MPS
	and quartet counters.
	Added a dummy calculation routine for the cavity BPMs.
	Made a routine to be used for runs before
	11233, which generates fake MPS and quartet counters.  Runs
	after this have hardware counters.  Modified the ntuples to
	include the new MPS and quartet variables.
	Modified AssignHelicity to use the new quartet phase counter.
	Added a routine to calulate the X and Y position at the
	target from stripline G0 and G0B.  Added decoding for the
	"energy measureing" BPM (stripline 3C12).
	In G0Hists, modified histograms for X and Y position at the
	target, energy, and helicity.
     *  The events before the first new quartet in a run are discarded.
	Added a interactive routine to set the French configuration
	mode in the realtime analysis.
     *  Minor changes to implement the French "back" scintillator
	when in "Front or Back" or "Back only" modes.
     *  Implemented SDMCH electronics decoding from the prototype
	code written by Dominique.
	At the run end, the summary of the SDMCH data is written to
	$G0SCRATCH/tmp/DMCH.<run#>.log (including the DAC values).
	Renamed the routines for the missing DMCH boards.
     *  Major changes in the helicity predictor to allow the event
	sampling to be non-continuous, such as when running in the
	realtime mode.
	The realtime code now does asymmetry calculations.
     *  Added command line flags to disable the asymmetry calculation
	of either Beam, NA, or French, or any combination.
     *	Added a command line flag to analyze single run segments.
     *  Minor midifications to g0acm to allow "smoothing"
	of the data.
     *  Changed the BPM names in the beam_nt and qrt_nt ntuples
	from names like "X01" to "x_G0"; all of the BPM names
	now refer to the device names.
     *  Changed the sign of x (BPM) calculation, so that it is consistent
        with the EPICS calculation.
     *  Minor bug fixes in GetReplayParams nad the helicity predictor.
     *  Modified TEtClientBuffer to make the functions "codaOpen" public.
	This permits a user to specify a non-standard ET station name for
	their version of the analyzer.
	If there are multiple analyzers all connected to the same station,
	they will share the events (two analyzers each get 50% of the events,
	etc.); so by specifying a unique station name, a process can be sure
	of getting as many events as it can take.
     *  Changes to allow the calculation of asymmetries from the SDMCH
	scaler data and for the NA direct scaler data.
	Added a histogram input file, to specify at run-time what the
	histogram properties should be.
     *  Fixed a bug in Helicity.C in the calculation of running average
	of the quartet yield. Fixed a bug in the quartet ntuple filling,
	where we rebin French timebin asymmetry to 1ns bins, which had a
	mistake in the weighted sum.
     *  Modified the retrieving of calibration data from gzero database
	according to the updated gzero database structure.
	Modified WriteToSQLDB() according to the updated gzero database 
	structure. At the end of the replay of each run, this routine is
	called and we fill the measurement table and beam table with 
	asymmetry and yield records of the time-encoding and beam data.
	Implemented the reading of cut table from the database. Also added a 
	text cut file g0cuts.in in prminput/ for non-database access. 
	Resurrected G0Cuts class to store cuts name, mininum and maximum, 
	etc.  
     *  Updated the MySQL query translator. This interface should now
	be ready to put into the client programs.
	Up to this stage the writing of gzero mysql database (parity
	analysis part) via g0analysis is in shape. The database (on hallcl7)
	also servers as the inputs (calibration, cuts, etc) of the replay.
     *  Updated the cavity BPM/BCM map.
     *  Updated the Meantimer->Scaler mapping to include all 16 NA detector
	rings.
     *  Implemented with Guillaume his deadtime correction routine for DMCH.
	MT deadtime was also corrected according to ~32 ns intrinsic deadtime 
	for left-right coincidence. DNL correction of the French electronics 
	is performed once at the end of a run.
     *  Updated the decoding to adapt one more DMA status word in each DMCH
	board. DMCH Alerts counters are written to the rate file, which
	"g0ratemonitor" reads.
     *  Updated the decoding of TS0 to accomodate the sub-bank of trigger
	supervisor data buffer. Coded a "wrapper" class G0TriggerSupervisor 
	derived from G0BeamMonitor, etc. This gives a more natural separation 
	between data of different sub-banks in the TS crate. Jason wrote a 
	G060HzPhaseMonitor class (also a base class of G0TriggerSupervisor) to 
	handle the 60Hz phase monitoring ADCs.
     *  More command line switches and flags:
	--tscfg <time>:	        selects between the TS config file with
				different time stamp
	--missingdmch <time>:	selects missing dmch board files by time stamp
	--tof <raw/corr>:	filling raw (no cuts) or corrected (cuts &
				DT corrected) ToF
	--french_dt <on/off>:	switch for French Deadtime correction
	--french_dnl <on/off>:	switch for French timebin DNL correction
	--tgthists:		flag for target boiling histograms
	--fbnt_withcuts <>:	fastbus ntupling with F&B coincidence cut
	--caesura <time>:	selects DMCH NPN cut off file by time stamp
	--dmchbinwidths <run#>:	selects DMCH timebin width from calibration
				run #
     *  To avoid confusion of the growing command line switches, implemented a 
	default switch file, Analysis/prminput/replayflags.in. But command
	line switches will override the flags in the switch file.
	HELP switch "-help" will now only give a very small list of replay
	flags for normal shift takers. "-helphelp" will print the full list.
     *  According to the scheme of fastbus setup file look up table, 
	implemented a lookup file, Analysis/prminput/inputfile_lookup.in, 
	for all the input files with timestamps (or other ID like "file.ID") 
	vs applicable run ranges, so that g0analysis will look up this file 
	and determine which ones are the right input files to take, based 
	on the run number which is being replayed.
     *  Renamed ROOT files to G0Scaler.XXXX.root, G0Fastbus.XXXX.root, 
	G0Full.XXXX.root, etc, to distinguish between different replay types.
     *  Added LUMI and HALO beam scaler channels into beamntuple.
     *  Fixed one bug about the histogramming of NA and French buddies.
     *	Modified the G0CodaEvent to be able to properly recognize events
	during 120 Hz running.  There is no analysis yet set up for
	120 Hz events, however.
     *	Changed the way the ROOT file is saved at the end of the analyzer to
	eliminate the duplicated ntuples in the file.
     *	Entered the "sub-bank" decoding mapping file for the runs during
	the testing and after run 13331.
     *	Modified beam.in.10-23-02 to include the HALO and lUMI channels.
	Updated the pedestals of BCMs and C1 channels of two cavity monitors 
	using Gary's new calibration values.
	Added the asymmetry calculation of C1 channel of cavity 1 and 2.
	Put the asymmetries into the quartet ntuple.
     *	Minor changes in the error checking when reading the subbank structure
	of TS0.
     *	Added NA LTD error histogram (32 bins).
     *	Resurrected "+- bit error" output file.
     *	Modified "--tof <>" flag to "--tof <raw_nocuts/raw_cuts/corr>".
     *	Fixed a "dividing by zero" bug in NA deadtime correction routine.
     *	Fixed a bug in G0RealTime so that the helicity counters are only 
	checked for 30 Hz events.
	Small fix to a comment line in the TS config file.
     *	Fixed the problem of getting the DMCH configuration from command 
	line input in realtime.
	Fixed a bug in SDMCH decoding with Dominique.
     *	Moved GMS_mask and E120 related methods from G0CodaEvent to 
	G0TriggerSuppervisor class.
	Fixed the decoding problem of gms_mask and e120 according to the
	new TS0 configuration.
     *	Beginning with run 13830, six words were added to sub-bank #1 of TS0.
	The beam lookup file "beam.in.11-07-02" has the new words listed as
	"spare".  THe input lookup file has been modifed to use the new file
	from now on.
     *	Changed BCM3 to UNSER in the beam input file, beam.in.11-07-02.
     *	Modified the entry names in mps_ntuple,renaming x01, x02 ... to 
	x_G0, x_G0B ...
     *	Added helicity correlated difference of LUMI/charge into quartet
	ntuple.  Added Halo histograms into realtime histograms list.
     *	In FiB buddy mode, put the DMCH buddy spectra into the
	"fr_back_tofXXX" histograms.
     *	Updated the BCM gain and offset in g0analysis according to g0log #7005.
     *	We did the BPM pedestal calibration before and after we swapped the
	bi-polar and uni-polar V/F. Two individual input files.
	The inputfile_lookup.in has been updated to use
	beam.in.11-13-02.swappedV-F for later runs.
     *	Read in lumi pedestals from the input file.
	Updated all of the older beam input files to include the
	offset and scale parameters for the Lumi and Halo monitors.
     *	Added a new time bin width calibration file for DMCH based on
	run 13978. We will use this file as default from now on.
     *	Implemented the DMCH deadtime correction using differential buddy
	counters suggested by Domi.
	Added a command line flag to switch between French deadtime correction
	methods.
	Added switch "--na_dt <on/off>" to turn on/off NA deadtime correction.
     *	Modified "detBinRange.in" in $G0ANALYSIS/Analysis/prminput to reflect 
	the +-2 sigma cuts of proton bins of the ToF spectrum found by Benoit.
     *	Fixed a bug in NA and French normalization routines.
     *	Fixed a bug in LUMI asymmetry calculation routines.
     *	Modified the scale of proton asymmetry histogram in "g0hists.in" in
	$G0ANALYSIS/Analysis/prminput/.
     *	A minor modification concerning the SDMCH alerts ----  
	Added DMCH board <-> Octant, Detector mapping for the "alerts"
	messages in DMCH.XXXX.log file.


G0Analysis Release 0.12
     Paul King; 2003/01/30, 15:35.

     This release includes all of the modifications made since the beginning
     of the G0 commissioning run.

     Changes since release 0.11:
     ===========================
     *  Added helicity entry into the fastbus ntuples. Implemented the 
	input file "tdc_cal.dat.[]" for TDC calibration (for now just take
	1 for all of them). Modified the range of the histograms 
	(ADC vs TDC) in realtime so that NA ADC hits will show up.
     *  Added NA LTD bit error debugging histograms into the "tof" 
	directory to replace the old "bit error" output file. Made 
	"raw_nocuts" as the default option for "--tof" flag (filling 
	ToF histograms with the raw yield before any cuts on beam).
     *  Added the asymmetry calculated from 4-wire sum of G0 and G0B BPMs 
	into the quartet ntuple.
     *  Updated the proton peak cuts after the shift of YO.  Added some
	locally modified input files on g0ana into CVS. 
     *  Fixed a bug in the MPS ntuple filling routine. The location where 
	it was called in the main routine was wrong --- the normalized yield 
	which gets written into the ntuple is always lagged by one MPS. Some 
	of the runs has to be re-analyzed if you want to reconstruct 
	the quartet sequence from the MPS ntuple.
     *  Added more input files for the proton time bins (calibrated at 
	different points). The analyzer will choose the appropriate one 
	by the run number according to the entries in "inputfile_lookup.in".
     *  Modify BCM2 pedestals. The BCM2 pedestal calibration calibration 
	which was done in the early November did not reflect the real 
	pedestal of BCM (off by about 200 channels).
     *  Fixed a bug in G0FrenchDetector::Map_Scalers_Directs() routine.
	Somehow the initialization section of this routine will overwrite 
	the memory location of the high end of the proton cut of octant 8 
	detector 16.
	Altered the loops in a safer way and seemed have fixed the problem.
     *  Found a bug in the beam ntuple filling routine; the MPS counter 
	was being filled with the software event counter instead of the 
	hardware MPS counter.
	This caused discrepancies when running with interleaved Fastbus 
	events, since the event counter was incremented multiple times 
	between MPS events.
	The beam ntuple variable "imps" now is filled from the MPS hardware 
	counter.
     *  Some development concerning G0 database.
	- Now make a little use G0ControlEvent to store start time and 
	  stop time into run table.
	- Expanded the run table to include actual number of 30Hz, 120Hz
	  and fastbus events of the run. Also added "target" column. Created 
	  electronics table to store the configuration information. Modified 
	  Analysis table to store the "good" 30Hz, 120Hz and fastbus events.
	  These numbers could be different
	- Modified WriteToSQLDB() to fill run, analysis and electronics table
	  properly.
     *  For the offline histograms, now support building the individual 
	detector proton asymmetry histogram (not the ring sum) for more than 
	one rings. Modify the "qrthists" flag in "replayflags.in" to achieve 
	this.  The syntax is the same as fastbus ntuples (ring number 
	separated by ":" or "all").
     *  Added to more flags "--binrange <>" and "--switchfile <>" into 
	"-helphelp". Forgot to this in the previous version.
     *  Added a selection of charge normalizer in G0BeamMonitor::Comp_charge().
	So choose BCM1 when beam current is higher than 30uA otherwise BCM2.
     *  If no histograming or ntupling flags are specified, no rootfile will 
	be generated. Got rid of "G0Empty.xxxx.root".
     *  Modified the fastbus ntuple to store some 30Hz info. Added two 
	branches in the fastbus tree --- "q_targ" and "imps". 
	"q_targ" is the charge on the target per MPS and "imps" is the 
	hardware MPS counter. A similar entry, "misc.macro" is the CODA 
	event counter for historical reasons.
     *  An attempt to make the realtime analyzer un-interrupted by the 
	end-run of DAQ.
	Minor modifications in GetPhysicsEvent() so that if evRead() return 
	non-zero value and the analyzer is not "g0analysis", we wait for 
	5 sec and try again instead of quitting the program. Also made 
	GetPhysicsEvent() an integer function instead of a void.
	Minor modification in G0Realtime.C. Only do the decoding stuff 
	and the subsequent calculation and histogramming of a CODA buffer 
	if GetPhysicsEvent return 1.
     *  Modified the normalization of lumis to switch between BCM1 and BCM2 
	according to the beam current, just as I did for the FPDs.
     *  Additional minor bug fixes.
     *  Addition of new parameter files covering the commissioning running
	period.
     *  Addition of some minor changes in G0Globals for the G0ACM, and
	some changes in the G0ACM routines.  These were copied from
	the working copy of the analysis used during the commissioning
	run.


G0Analysis Release 0.13
     Paul King; 2003/03/20, 10:35.

     Interim release containg a few small changes; the modifications to
     include K. Nakahara's 6 parameter regression are expected shortly,
     so this release will freeze the condition before that is added.

     Changes since release 0.12:
     ===========================
     *  Modified the analyzer to improve the EPICS and control event decoding.
	Individual EPICS events are decoded, and a running mean, sigma, minimum
	value and maximum value are calculated for each recorded variable.
     *  Fixed a minor compilation error found using the Sun compiler;
	four member constants were initalized in the header file instead of
	in the source file.
     *  The GNUmakefile was modified to add some error checking so that the
	cleaning or compilation will not take place if the environment
	variables are not set consistently; this will help to reduce the
	occurance of accidental deletion of the "group" version on the
	common accounts, and deletion of files in the home directory of
	individual users due to incomplete environment variable settings.
	The ${G0ANALYSIS} variable must agree with the directory from which
	the make is launched, and the ${G0BIN} and ${G0LIB} directories must
	be subdirectories of ${G0ANALYSIS}.
     *  Installation of the modifications of the G0FrenchDetector class made
	by Guillaume to include the French detector single rates in the
	deadtime correction.
     *  Applied Guillaume's bug fixing in French DNL correction routine in 
	G0Hists.C.
     *  Took BCM1 to do the normalization throughout --- "q_targ = charge[0]".


G0Analysis Release 0.14
     Paul King; 2003/05/21, 19:39.

     This is the production release for the "second pass" analysis.

     Changes since release 0.13:
     ===========================
     *	Implemented code to write EPICS variables directly into v0.7
	of the G0 database.
     *	Added a buddy ntuple.

     *	Several changes:
	1. Implemented Kaz's linear regression class
	    G0LinearRegression.  This class calculates the detector
	    yield vs beam parameter slopes in the first  pass of the
	    replay. In the 2nd pass the slope correction will be
	    applied.  Added the newly calibrated x_targ, y_targ,
	    theta_x, theta_y, engery, q_targ in G0BeamMonitor class.
	2. Modifications on G0Scintillator class:
	    a) Added "proton","pion","inelastics","bkg" and "total" data
	    member in G0Scintillator to ease Kaz's yield correction. Also
	    defined the corresponding proton_low, proton_high, etc, in
	    G0Scintillator to store the PID cuts. Removed the obsolete
	    window_low, window_high from G0NADetector and
	    G0FrenchDetector. Removed the PID weight calculation in
	    the histogramming and ntupling routine.
	    b) Moved the DNL timebin width fWidth[] from individual NA and
	    French detector classes into G0Scintillator.
	    c) As an effort of code clean-up, defined the overloaded 
	    operators "+", "-", "=", "+=", Divide()" "MeanValue()", "Sigma()",
	    "Normalize()" , etc at this level. These operators are
	    involved in the asymmetry and  yield calculation by
	    G0NADetector and G0FrenchDetector objects.
	    d) Removed obsolete data structures from G0Scintillator.
	3. Implemented three more replay flags for Kaz's linear
	    regression analysis: "--calcslope", calculates the slopes
	    during the replay "--linregcorr", applies the slope
	    corrections to the detector yield "--writelinregprms",
	    writes the slopes to an output file.  Updated the flags
	    for the database interface: "--cut_id <>", to specify the
	    cut_id "--calib_id <>" to specify the calibration_id
	    "--dt_rates_id <>" to specify the deadtime_rate_id.
	4. Moved french_config structure from G0CodaEvent to
	    G0FrenchDetector.  Also move the decoding routines into
	    G0FrenchDetector. Added a structure  to store all the DAC
	    settings for later filling into the database.
	5. As Julie previously suggested, removed the timebin by
	    timebin entries  of the yield and asymmetry from the
	    quartet ntuple. Replace them with the  "proton" yield and
	    asymmetry for each detector (8*16).
	6. Added database input retrieval hooks for the following
	    parameter sets:
	    a) PID cuts (G0NADetector, G0FrenchDetector)
	    b) DNL (G0NADetector, G0FrenchDetector)
	    c) French NPN cutoffs (G0FrenchDetector)
	    d) Single rates needed for the deadtime correction
	       (NA: tartan rates, French: MT and CFD single rates)
	    e) Length of the deadtimes (G0NADetector, G0FrenchDetector) 
	    f) beam monitor calibration constants (G0BeamMonitor) 
	    g) linear regression slopes and means (G0LinearRegression) 
	    h) fastbus TDC calibration (G0Fastbus) 
	    i) global cuts (on beam current, etc.) (G0Cuts)
	7. Updated the database filling routines WriteToSQLDB()
	    etc 
	    a) updated measurement table filling. 
	    b) updated beam table filling.  
	    c) slopes table filling (new).  
	    d) french electronics tables filling (new).
	8. Modifications in G0SQL class: 
	    a) Besides detectors and monitors tables, also load particles 
	       table  and slope_types table in the c-tors. 
	    b) Coded two selection routines to retrieve the default 
	       calibration_id  and deadtime_rates_id (SelectCalibID() 
	       and SelectDTRatesID()) if nothing  is specified. Otherwise 
	       take the values from command line input 
	    c) Implement the filling of flag columns in the analysis 
	       table during  the replay.
     *	Corrected table names for magnet variables in 
	G0EPICSEvent::DefineEPICSVariables().
     *	A few bugs were found and fixed.
	1. I forgot to put the "total" key word parsing in
	   G0FrenchDetector.
	2. The new FillFrenchCfgBuffer() was offset by two words.
	3. NA DNL file was not implemented yet so the time bin widths
	   were zeros.  Consequently the deadtime correction created
	   "nan". So took a dummy  routine to assign bin widths ---
	   6.5ns, 1ns, 1ns, ... instead.
	4. This bug came from the (newly implemented) overloaded
	   G0Scintillator operators. The copy c-tor of G0Scintillator
	   took kTimeBin (a global non-const)  to resize the timebin
	   vector and kTimeBin changes from 24 to 128  globally in the
	   G0NADetector and G0FrenchDetector c-tors to take into  account
	   the difference between NA and French. This trick  had bad
	   effects when the "+" and "-" operators of G0Scintillator  were
	   called to do the asymmetry calculation, since kTimeBin was
	   always  reset to 24 after the French c-tor. So defined an
	   internal variable  fNumberOfTimeBins in G0Scintillator so that
	   when the copy c-tor is called,  it knows the right size to
	   ensure the asymmetry calculation be performed  on ALL the time
	   bins.
	5. In this new version I stick to a Divide() function when
	   doing the  dividing between numbers. Set the output value to
	   +100,000.0, if the  denominator = 0, to be consistent with the
	   old engine.
	6. Cross-check the asymmetry output file with the one produced
	   by the  old engine (02-18-2003 version). Differences were
	   found on the French timebins.  I just realized in this new
	   version we have had Guillaume's modification of  the French
	   deadtime correction. Put this into the old engine and ran the
	   check again. indeed it solved the discrepancy.
	7. Small fixs of the headers of TEtClientBuffer.C and
	   G0OnlineEvent.C to  quiet gcc 3.2.

     *	Found an error in the decoding of the French configuration
	event when the event limits flag was used to skip the file segment
	containing the event (such as using "-e 470000:490000").
	Resolved the error by moving the processing of the French configuration
	event and the EPICS events outside of the test on the event number, 
	so that these events are processed even if they are not within the 
	event limits.
     *	Calibrations:
	1. Put Aamer's new BCM calibrations into the relevant pedestal files 
	   of January runs.
	2. Added the "q_targ" entry in the global cuts (g0cuts.in).
	3. Added Retief's new PID calibrations files (detBinRange.XXX.work) 
	   into the repository. These new calibrations should be applied to 
	   the January runs. 
	   The "proton" and "pion" cuts were carefully checked. Defined 
	   "inelastics" window between pion_high and proton_low, and 
	   "background" window between proton_high and the end of ToF. 
	   Modified inputfile_lookup.in to reflect this. I will talk to 
	   Paul to remove the old calibration files.
	4. Modified the default NA deadtime correction flag to "dog_only" 
	   in replayflags.in.

     *	More modifications related to the upcoming production replay.
	1. Bugs fixed during the database filling test (most in sql query 
	   string compositions)
	   a) NA&French:: LoadPIDBins(), LoadDNL().
	   b) Commented NA::LoadDNL(sql). No final calibration values 
	      available yet.
	      In NA class, added array structures fDTRateFitP1[] and 
	      fDTRateFitP2[] to store the quadratic fit parameters of 
	      tartan rates vs beam current. 
	      Renamed NA&French::LoadDTValues(sql) to LoadDTPrms(sql) 
	      and implemented the retrieving of quadratic fit parameters 
	      in NA::LoadDTPrms().  Commented NA&French->LoadDTRates() 
	      and sql->SelectDTRatesID() in GetInputData(). 
	   c) To follow CMU convention, modified tartan axis definition 
	      in NA class in terms of left and right instead of front and back.
	   d) Added LoadTDCCalibPrms() in fastbus class to retrieve the 
	      gain and offset of the TDCs from the database. To keep the 
	      backward compatibility, still fill the fastbus ntuples with 
	      the raw data. Might change it later.
	   e) Modified FillAnalysisTable() and FillRunTable in the sql 
	      class for properly filling. In the main code, added 
	      sql->FillRunTable().
	   f) Small bugs fixed in WriteToBeamTable() and 
	      WriteToFrenchElectronicsTable() and G0EPICS::WriteToDatabase().
	   g) Modified LoadLinRegPrms() method in the linear correction 
	      class to properly retrieve the slopes and means from the 
	      1st pass replay. Moved sql->StoreReplayFlags() in 
	      GetInputData() to ensure the knowledge of all the relevant 
	      flags before sql->FillAnalysisTable() is called.
	2. Implemented Silviu's beam trip cut (we throw away the first 
	   2000 MPS after the beam recovers from a trip) in G0BeamMonitors 
	   and PassedCuts.
	3. Expanded the NA deadtime correction option list ("--na_dt <>") 
	   to "off", "70ns", "dog_singles" and "dog_only" to switch between 
	   different correction approaches.

     *	The function: SlopeError()  was changed to correctly include all 
	the off diagonal matrix elements in the covariance matrix for the 
	error calculation.
     *	Another bug fix from Kaz in G0LinearRegression.
	This corrects the array indices for the variable "xvarsum" (should be
	one-dimensional, but was indexed as multi-dimensional).
     *	Added a "good_event" flag for the fastbus events. This flag is set 
	by the "goodness" of the most recent 30Hz event so effectively the 
	beam current and beam trip cuts will apply to those fastbus events.
     *	Added a constructor method for G0SQL which takes the MySQL server 
	host name, user name, and database name as string arguements.  
	This was done to simplify development of scripts which extract data 
	from the database.
	The default contructor now calles this new constructor with empty 
	strings as the arguements.
     *	Updated the G0 sql query translator according to the new database 
	structure in G0SQL class.
     *	Added Damon's new G0Database lookup file, v0.7 into the repository.  
     *	Fixed bug that caused 'sms' and 'target' tables in the database to 
	be filled twice.
     *	Added code to fill polarized source tables from EPICS data stream.
     *	Modified the database lookup table to indicate the calibration_id 
	and cut_id variables are included in the analysis table.

     *	Squashed the following bugs:
	1)  Column name in 'measurement' table of G0 database changed 
	    from 'measurement_type' to 'measurement_type_id'.  Corrected 
	    code in G0SQL::Translate() appropriately.
	2)  G0 database table relationships in lookup table are order
	    sensitive, i.e. run.run_id:analysis.run_id appears in the
	    table but analysis.run_id:run.run_id does not.  Corrected
	    code in G0SQL::GetTableRelations() appropriately.
	3)  Fixed G0SQLResult::GetArrayF() to return NULL if the array
	    corresponding to the desired column could not be found.

	Implemented the following feature:   G0SQLResult can be used
	to store query results that were not generated from SQL
	created by G0SQL::Translate().  Added code to G0SQL::Process()
	to get column names from the returned TSQLResult in this case.

     *	More changes for deadtime:
	1. Implemented the retrieving of DT correction parameters in 
	   G0FB_deadtime class. Basically copied LoadDTPrms() method 
	   from G0NADetector.
	2. Added WriteDTRatesTables() routine in WriteToSQLDB to 
	   fill the tartan rates into the database at the end of the replay.
        3. Fixed the unit conversion of NA and French asymmetries before 
	   writing to the database. Coded the base method in G0Scintillator 
	   instead.
	4. Fixed one more bug in G0SQL::GetTableRelations().

     *	Squashed bug that caused all of the French TOF results to be 
	written into octant 7 detector 16 in the database.
     *	Added G0SQLResult::GetArrayName(UInt_t index) in order to retrieve 
	the array name for a given column index.

     *	Modifications to allow the use of G0SQL and G0SQLResult classes 
	at the ROOT command line and in interpreted scripts.
	1)  Added ClassImp and ClassDef macros to header and implementation 
	    files of G0SQL and G0SQLResult in order to get ROOT dictionary 
	    generation.
        2)  Added LinkDef files for G0SQL and G0SQLResult to control 
	    dictionary generation.
	3)  The dictionary generation (rootcint) really does not like 
	    vector<> in G0SQL.  Used __CINT__ macro to to prevent 
	    vector<> declarations and functions from being included 
	    in the dictionary.
	4)  ROOT interpreter has a lot of problems with the C++ string 
	    class.  Have overloaded various G0SQL and G0SQLResult member 
	    functions that took or returned string to make Char_t* versions.  
	    Important Note:  It is necessary to delete the pointer returned 
	    by G0SQL::Translate(const Char_t* selection, const Char_t* cuts).
	5)  As a consequence of 4) the G0SQL constructors now call an Init 
	    function to load up lookup tables and the like.

     *	Changed member variable sql_query to fSQL_Query. Store the whole 
	SQL string into it at the end of ComposeQuery().
     *	Dominique's fix of the French electronics configuration struct.
     *	Added the filling of "Buddy marker delays" into the G0 database.
     *	Some database filling tests and bug fixing before the production
	replay.
	1) Some modifications according to the recent change of 
	   "tartan_axis_id".
	   Added a tartan_axis table loader at the initilization and an 
	   id getter.
	   Fixed G0FB_deadtime::LoadDTPrms() and WriteToDTRatesTable() 
	   routine for NA deadtime correction accordingly. 
	2) Fixed a small big in updating the polarized_source table.
     *	Some cleanups of G0SQL class definition. Moved all the internal 
	functions related to the translator into "private". 
     *	Modifications from Angela to fix a bug regarding the 
	namefile of the tartan table written in the $G0SCRATCH/tmp directory.


G0Analysis Release 0.15
     Paul King; 2003/05/29, 09:45

     Incorporates several bug fixes found during the first stage of
     the second pass analysis.

     Changes since release 0.14:
     ===========================
     *  Put the beam cuts we are using in the full replay into 
	the analysis CVS, as g0cuts.in.1steng. This is associated with 
	cut_id 2 in the database.
     *  Some minor changes concerning the replay engine.
	a) In updating the run and analysis table at the end of the
	   replay, now  fill "start_time" into the run table, even the
	   END_EVENT is not reached  in the run. Added the inserting the
	   "number_good_quartets" into the run_table.  
	b) Updated the default replay flag file, replay_flags.in 
	c) Modified WriteToMeasurementTable routine, so that in the 
	   pass (b), where  the slope correction is applied, only the 
	   yields and asymmetries of the "pids"  will be filled in to the
	   measurement table.  
	d) Swapped the location of LinRegCorr() and FillMPSNtuple() 
	   in the main routine.
	e) A minor bug fixed in GetReplayPrms().  
	f) Fixed a bug in G0SQL::StoreReplayFlags(), in which the flags 
	   of the  current replay are stored.  "Write" the "last_event" 
	   as an unsigned int to be consistent with the data type in 
	   the database.
     *  Changes related to the SQL query translator.
	a) Fixed a bug in translating the selection strings,
	   when"(err)" is  involved.  
	b) Implemented the "group by" selection syntax for the measurement
	   table.  So now "[y,a]_detN[pid](err)" will retreive the "pid" 
	   values(errors) from the  measurement table, group by the detector 
	   number (and run_number if applicable).  
	   Similar syntax: "[y,a]_octN[pid](err)","[y,a]_xxxbinN(err)" 
	   is also  supported.
     *  Modified the default flags in replayflags.in
     *  Fixed bug related to the fortran to c array indices conversion 
	in the retrieval of the correlation slopes from the database. 
	The correlation slopes AFTER the linear regression correction were 
	verified close to zero.
     *   Various fixes for the MySQL interface.
	 a) Added timebin_id into the field name look up table. Fixed an 
	    error in the relation lookup table for "slopes" vs "slope_types".
	 b) Now supports quotes text, e.g. 'xxx', "xxx" in the cut string 
	    of the SQL query translator. 
	    Also supports [table_alias].[field_name] syntax in the cut string.


G0Analysis Release 0.16
     Paul King; 2003/06/03, 19:25

     The principle bug fix is in the NA deadtime correction routines
     used to extract the "tartan table" rates from the database records.
     This bug resulted in the zeroing of the deadtime corrections in
     the previous full pass of the analysis.

     Changes since release 0.15:
     ===========================
     *  A "nasty" bug fixed related to the "70ns" NA deadtime correction.
	After the retrieval of the tartan rates fit parameters, a
	small calculation routine  is needed go from the 4x4 tartan rates to
	CFD, MT, coin rates fit  parameters. This piece of calculation code
	was present in the input file  routine, but was missing in the
	database reading routines. This caused  the failure of performing the
	NA DT correction properly with the DB I/O.  Strip it out as a small
	subroutine.
     *  Reinterpret ToF PID cuts definitions [low,high] to be inclusive of 
	the cut edges. So [15,18] bin cut represents 15,16,17,18. Modified 
	G0Scintillator::CalcPIDSum().
     *  Added the "!=" translation in the SQL cuts translator.



G0Analysis Release 0.17
     Paul King; 2003/06/06, 10:45

     The principle bug fixes are to: the "beam trip cut" and the
     regression analysis routines.  The fix to the "beam trip cut"
     is so that runs which do not have the beam on at the beginning
     of the run are handled properly.  The fix to the regression
     analysis corrected several errors which corrupted the slopes
     calculated in the analyser.


     Changes since release 0.16:
     ===========================
     *  It was discovered that rolling averaged PID yields were unreasonably 
	large, for a few runs where the beam was off at the beginning.

	I forgot to tag the quartet as bad, if the MPS passes the beam
	current cut  but failed Sivliu's trip cut (2000 MPS after the beam
	recovery). For the  2001 MPS after the beam recovery, say if it is
	the 4th MPS in a quartet,  the asymmetry and the yield would be
	calcuated, since the quartet was  mistaken as good. But the first
	3 slots had some unknown values, which  created crazy results ...
    
	So the quartet flag is explicitly set to false whenever a MPS cut
	is failed.  Also put the safety belt in PassedCuts()
	routine. Modified the c-tor of  G0Scintillator so that it
	explicitly sets all the initial values to zero.
     *  Fixed several bugs with Kaz in the linear regression code. It was 
	discovered that although the output correlation slopes
	appeared to be  zero after the correction was applied, the
	true correlation still existed  and made the asymmetry
	senseless. Most of the issues were due to the  incompatibity
	between g0analysis and the original root script of the
	regression analysis. We removed the incorrectly incremented
	internal  event counter from the octant and detector
	loop. Also moved the  incorrected running summing of the beam
	parameters outside the loop.  Changed the way the rolling sum
	is calculated to MPS-wise, instead  of quartet-wise. Otherwise
	this had a very pethogical flaw, so that if the  replay ends
	at the middle of the quartet the slope will NOT be correctly
	calculated.
     *  Put a quick fix in G0CodaEvent::FillFrenchCfgBuffer(), so that if 
	the bank frag length appears zero, reassign it to 389. This is
	a  hack to  get through a couple of runs, where the French config
	event length appeared to  be zero and caused the segv in the replay.
     *  Added a function G0ReplayPrms::SetThisRun() to set a internal 
	variable to the current run number.
     *  Changed the default flags for the replay database host, database 
	name, username, etc according to the new DB on wmg0.
     *  Just commented a couple of debugging output in G0LinearRegression.


G0Analysis Release 0.18
     Paul King; 2003/06/06, 16:15

     There were some changes in the database access routines for filling
     and extracting data.

     Changes since release 0.17:
     ===========================
     *  Modified G0EPICSEvent::WriteDatabase() to handled cases where
	"static" EPICS variables change in the middle of a run.  The
	static variables include target position, insertable half wave
	plate state, and charge feedback state.  If the target position
	changes during the run, then the position will be set to 'u' in
	the database.  If IHWP or charge feedback state changes, then the
	value in the database will be NULL.
     *  Raw SQL will no longer be printed to standard out by default 
	before the query is submitted to the database.
     *  Defined two ERR constants in Math_Func, InvalidAsym = 100,000. and 
	InvalidSigma = -100,000. The former is to indicate divide-by-zero 
	error when calculation asymmetry, the other is to eliminate the 
	pathological case where  <x*x>-<x>*<x> < 0, due to the rounding error.
     *  When filling the DB with the FPD, beam and correlation slopes values, 
	check whether the numbers indicates the record is invalid due to the 
	above two errors. If yes, fill NULL into the DB fields. Modified 
	G0SQL::Process() routine to handel NULL records in retrieval (the 
	row with any NULL record will be skipped).


G0Analysis Release 0.19
     Paul King; 2003/07/03, 11:45

     The principle change is the inclusion of Jason's 120 Hz analysis
     routines.

     Changes since release 0.18:
     ===========================
     *  I have modified the routine which fills the "na_rates" table so that it
	will write out the rate information for all eight octants.  So the 
	"na_rates" table will have the rates calculated using the NA technique
	for all eight octants; perhaps we will rename this table in a future
	version of the database.
	The problem is that the array of the "measured rates calulated using
	the Fastbus data" is valid for all eight octants, and not just the 
	four NA octants.  This array had been accessed by looping over the 
	first four octants, getting the detector ID, and then outputing the 
	data from the array.
	So, the "na_rates" table as it exists in the database right now has the
	rate data for octants 1, 2, 3, and 4.  There is not a problem with the
	detector ID in the "na_rates" table.
     *  Fixed bug that caused duplicate entries in the 'polarized_source', 
	'sms', and 'target' tables (and their subsidiary data tables).  Now 
	the database will be checked prior to filling the tables to make sure 
	the data is not already in there.
     *  Added Jason's 120 Hz analysis routines:


	The revision adds 120Hz analysis to g0analysis, defining the
	new flag --120Hz.
	2 new features are also added :
	- beam offsets are now normalized against clock, which takes into
	  account small clock counts variations (typically +/- 1 over
	  ~133330 in 30Hz mode and over ~33330).
	- Device used for normalization can be specified at runtime.
	  Corresponding flag is :  --normcharge  <BCM1/BCM2/UNSER/LUMI1
	  /LUMI2/LUMI3/LUMI4>
	- Beam cuts can be inhibited, allowing to decode test runs with no
	  beam data. Corresponding flag is --nobeamcuts

	In order to introduce 120Hz data handling without affecting
	the code structure, the 30Hz objects are replicated and
	switched regularly within the standard loop. Basically, all
	data structures are instanciated and initialized as usual.  If
	120Hz analysis is required then a G0120Hz object is created
	that adopts and replicates all the previous objects. Its main
	task in the main g0analysis loop is to keep track of the 120Hz
	counter and switch around the 120Hz replica accordingly. Such
	a scheme makes the whole G0 analysis code extensively reusable
	with minor compatibility issues as each 120Hz structure is
	handled as an independent 30Hz structure. Only minor non
	transversal changes had to be here and there (see bellow), the
	main of which are the special handling of the French data
	which are still acquired on an MPS per MPS basis and the
	addition of a series of new tests in the main G0Analysis file.

	The output is similar to that of the 30Hz code, with a few
	caveats: a few variables are added in the standard ntuples,
	namely phase60Hz, phase60HzSum and phase120Hz; phase60Hz is a
	float comprised between -pi and pi : it is the power line 60Hz
	phase as reconstructed from the 3 ADC values 60HzPhaseADC# in
	the beam ntuple.  phase60HSum is the same, but the
	reconstructed value increases ad infinitum and no modulo is
	applied any more.  phase120Hz is the 120Hz phase counter,
	ranging from 0 to 5 (0 and 5 have special meaning : 0 ==
	1+2+3+4 and 5 = 1-2+3-4) The standard histograms are
	distributed between directories "phase0" though "phase5" and
	their names are slightly modified, with the suffixes "_ph0"
	through "_ph5" appended. All the rest is preserved.  e.g. :
	tof/fr_tof201 -> phase#/tof/fr_tof201_ph#

	Change log :
	** Asymmetry.C : operator += and -= were declared
	   but not defined 
	** G060HzPhaseMonitor.h/G060HzPhaseMonitor.C :
	   Calibration method added, which is used to determine the
	   constants to reconstruct the power line 60Hz phase.   
	   Right now the latter constants are still hard-coded.
	** G0BeamMonitor.h/G0BeamMonitor.C : - all beam vector renamed
	   from "name" to "name_per_clock" for clarity - New parameter
	   NOMINAL_30Hz_CLOCK for pedestal normalisation; right now is a
	   constant : may have to be defined as a variable and be put in
	   the data base - New parameter fNormChargeIndex, defines which
	   charge is used as normalisation.  - operator += and -=
	   (expected by G0120Handler) - Clock computation separated from
	   charge computation in new method Comp_clock()    
	** G0Buddies.h/G0Buddies.C : added members
	   oversampledphase[4],phase60Hz[4], phase60HzSum[4];    
	** G0CodaEvent.h : Revised Is120HzMode() test method    
	** G0EPICSEvent.C : Changed test value.Contains(digit_regex) into
	   test value[0]>='0'&&value[0]<='9' (regular expressions
	   handling is too recent a feature in ROOT)    
	** G0FrenchDetector.h/G0FrenchDetector.C : - New method
	   Decode120Hz() to handle special French 120Hz data structure -
	   Method ResetValues() to prevent transmissions of non-zero
	   values from run to run (typically, after a config change,
	   where a subsystem wouldn't be read any more, for test reasons,
	   then the corresponding ROC CODA header are missing, but the
	   buffers were not reinitilaized)  
	** G0Hists.C : Naming scheme slightly altered to distinguish 
	   between 120Hz histograms    
	** G0NADetector.h/G0NADetector.C : - Method ResetValues() to
	   prevent transmissions of non-zero values from run to run
	   (typically, after a config change, where a subsystem wouldn't
	   be read any more, for test reasons, then the corresponding ROC
	   CODA header are missing, but the buffers were not
	   reinitilaized) - Defined operator += and -= which used to be
	   only declared (expected by class G0120HzHandler)    
	** G0Ntuples.h/G0Ntuples.C : 3 branches added to the 4 standard
	   ntuples : "phase120Hz","phase60Hz","phase60HzSum"    
	** G0ReplayPrms.h/G0ReplayPrms.C : - rootfile, fNoROOTFile are
	   now static (necessary in 120Hz as closure time is random with
	   regard to the 120Hz counter) - Flags (+ getters and setters)
	   added for 120Hz mode, charge normalisation selection and beam
	   cuts inhibitions  
	** G0TriggerSupervisor.h :
	- transfered 60Hz phase computation in method ProcessEvent()
	** GetInputData.C : added line
	   beam->SetNormChargeIndex(io->GetNormChargeIndex())    
	** Helicity.h/Helicity.C : explicit default copy ctor added    
	** GetReplayPrms.C : flags "--120Hz", "--nobeamcuts",
	   "--normcharge"  
	** MapScalerChan.C : scope of vectors changed,
	   to prevent memory effect in initialisation. They are also made
	   static to make sure changes don't propagate.  It required
	   distinguishing between counters kChannel and kChannelforBeam.
	   Each call to MapScalerChannel() now first resets the vectors.
	** beam.in.01-12-03/beam.in.01-13-03/beam.in.01-23-03/
	   beam.in.11-07-02/beam.in.11-13-02.swappedV-F/beam.in.11-21-02/
	   beam.in.11-21-02.swappedV-F
	   11th field of beam buffer is now defined as "Phase120Hz"
	   (used to be "Spare")  
	** G0Analysis.h/G0Analysis.C : -
	   Conditional instantiation of G0120Hz object.  - Various tests
	   on the 120Hz phase to keep the analysis in synch (see comments
	   therein) - Hard-coded reordering of French 120Hz data
	   (correction for the shifted Grey code that labels the
	   different 120Hz phases) - Several bypassing tests to prevent
	   unwanted effects such as beam cuts on a single 120Hz phase
	   (typically the "5th"), which destroys the synchronisation    
	** Added classes G0120Hz and G0120Handler : the latter is a
	   templated container class that refers to 6 copies of a
	   definite data structure. It expects for the data classes that
	   the operator =, +=, -= be defined. The former is a
	   "synchronising" class. It contains almost as many
	   G0120Handlers as there are data structures in normal 30Hz
	   analysis. Its task is to maintain a synched index which
	   defines uniquely the sets of buffers currently used in the
	   analysis (corresponding simply to the indexth elements in the
	   G0120Handlers).


G0Analysis Release 0.20
     Paul King; 2003/07/09, 11:00

     The principle change is the 10 sigma cuts we impose on the 
     yield in each timebin. If any time bin in a detectors fails this cut, 
     the whole detector will be thrown away from this particular MPS and 
     quartet, i.e. the calculation of yield correlations, asymmetries, etc 
     will be skipped. This is to remove the non-statistical fluctuation 
     we saw on the detector yields, and hence asymmetries and the slopes.

     Also, Guilliaume's new deadtime correction routine for the French
     detectors has been added.  A bug in the helicity correlated angle 
     difference calculation was corrected.  Included the hardware counters
     of the MPS and QRT in the database.


     Changes since release 0.19:
     ===========================
     *  Modified G0Scintillator class. For each G0Scintillator object, added 
	an internal flag fGood, and a good event(quartet) counter, nGood. 
	Various additions to the computation methods to calculate based on 
	the goodness flag and the internal counter(for running average, 
	etc ...).
     *  Kaz modified the linear regression routine accordingly. Added interal 
	event counters for individual detectors. Only calculates proton yield 
	slopes (and later applies the correction) when the detector goodness 
	flag is true. Due to cuts (and counters) on different detectors, the 
	6x6 matrices have to be computed for individual detector.
     *  Modified G0Cuts class. Defined a few G0Octant vectors to store the 10 
	sigma cuts in the form of octant[].scint[].dtimebin[]. 
     *  Added files CutsOnDetectors.{C,h}, which are a collection of functions
	involved in the yield cuts.
	-- TagBadDetIfErrorAlert() looks at the LTD error counters and SDMCH 
	   alerts to tag the bad scintillator accordingly. This is the 1st 
	   level protection.
	-- CalcBinYieldRMS(). Called at the the beginning of the replay. 
	   Sacrifice 100 "good" MPS to compute the mean and RMS of the 
	   normalized yield of each bin. After this the 10 sigma cuts 
	   around the mean is calculated for each bin.
	-- TagBadDetIfFailYieldCuts(). Actual implementation of the 10 sigma 
	   cuts on the timebin yield. If any bin fails, tag the whole 
	   detector as bad.
	-- ResetDetFlgs(). Resets fGood for each detectors at the beginning 
	   of a new MPS.
	-- TagBadQRT4EachDet(). If any of the 4 MPS is tagged bad for a 
	   certain detector, tag the quartet as bad for this detector as well. 
	-- ReportGoodQRTCounters(). Output the nGood counters for each 
	   detector.
     *  Parallel modifications in the G0NADetector, G0FrenchDetector classes
	for flagging and increment of internal counters. Similar modifications 
	in higher level Helicity and Asymmetry classes. 
     *  Misc. modifications CalcAsyms(), CalcYieldAve(). Also modified the 
	quartet histogramming and ntupling routines to check the goodness 
	flag before filling.
     *  Corresponding changes in the main routine. Moved the calls of Kaz's
	regression methods inside the quartet cuts. Merged against Jason's 
	recent mod. on 120Hz analysis. Added the command line switch,
	"--cutonyield".
     *  Put Guillaume's new French DT corrections in. This version utilize a 
	"dead" probability lookup file to modify the yield correction due to 
	the singles timebin by timebin, based on the singles ToF info from 
	fastbus. 
	Now the default correction routine! Added a redundant switch 
	"--french_dt npn_single_tof".
     *  Fixed a bug in the helicity correlated angle difference calculation. 
	This bug caused the beam angle differences not filled properly in 
	the DB during the 2nd pass replay. 
     *  Put the value of the hardware MPS and QRT counter into the run table 
	in DB. 
	Also put the value of the hardware MPS counter into fastbus 
	ntuple->misc.macro.


G0Analysis Release 0.21
     Paul King; 2003/07/11, 17:30

     Included some minor changes in the database filling routines related
     to the "10 sigma" yield cuts.

     Changes since release 0.20:
     ===========================
     *  Some minor fixes in the database filling routines; the changes are
	related to the 10sigma yield cuts.
	Added the filling of "cutonyield" flag in the analysis table.
	Added the filling of "n_qrt" detector by detector in the measurement
	table.
	Added the filling of "n_mps" detector by detector in the slope table.


G0Analysis Release 0.22
     Paul King; 2003/08/15, 17:30

     Several errors found related to the "10 sigma cuts", and the cuts on 
     LTD errors and DMCH alerts.  We now require a minimum number of good
     quartets to put a detector result into the database.
     Some corrections of the linear regression corrections were made.
     Made some modifications to the histograms in preparation for the
     next running period.

     Changes since release 0.21:
     ===========================
     *  After another full pass of the data replay was finished, I discovered 
	that half (64 out of 124) of the "good" asymmetry runs has zero good 
	quartets. The problem  was related to the cut on the NA error counters.
	The first 100 "good" MPS (meaning that no error or alerts for both 
	electronics) was sacrified to calculate the sigma of the yield,
	after which the 10 sigma cuts was imposed onto individual detectors. 
	But for those runs some NA error was on all the time, so we never got 
	to the 100 good MPS and thus threw away all the events for both 
	electronics ...
	I temporarily commented the test on errors and alerts in the 10 sigma
	cuts calculation.
     *  Also to save the size of the replay log file and optimize the speed,
	commented the error message outputs for errors and alerts.
     *  Found a problem related to the sqrt() operation (used extensively in 
	computing the errors). The operand of sqrt() tends to be negative 
	when the code tries to calcuate the error of a quantity with zero 
	good quartets (in principle 0-0 should give zero, but somehow this 
	small negative number is beyond the limit of double precision or 
	something).
	This becomes a feature after the new cuts we implemented on the NA 
	error counters. Numerous of detectors were cut throughout the runs 
	such that the number of good quartets are always zero, which led 
	to negative operand of sqrt() when calculating errors. In the end 
	at the DB filling routine, the code exited, since the DB doesn't 
	like "nan" to be filled.
	So added a wrapper function SQRT(x) in the Math_Func.h, such that it 
	returns a distinct big error number if x<0. Called this function 
	in the error calculation at various places.
     *  After discussing with Paul, we decided to stick to a single 
	"G0 invalid number" in all the calculation routines to clearly 
	distinguish the bad number from good. Upon filling the DB, do a 
	test on the filling value and fill NULL if the value = G0InvalidNumber

	The current defined user functions in Math_Func routines are:
	G0InvalidNumber = -10000000.
	Sigma(mean, meansq, n)
	SQRT(x)
	Divide(x,y)
	Product(x,y)
	The regression routine still need more cleanups. For now just make 
	sure to use user-defined SQRT() and Divide().
     *  Fixed the minima of the 10 sigma yield cuts to zero, if mean-10*rms <0.
	Also if rms is zero or G0InvalidNumber (dead detector indicator), 
	set both cuts to G0InvalidNumber, such that no events of this 
	detector can pass.
     *  Defined a cut on the minimum number of good quartets (by default
	1000). 
	If the quartet counter for a certain detector is less than this 
	value, no valid numbers (asymmetry, yield, slopes, etc) will be 
	filled into the DB. 
	Applied the same cuts in the regression routine to remove the frequent 
	singular matrix errors (if we don't have enough statistics the code 
	does not bother to caculate the slope at the end of the replay).
     *  More on command line flags:
	"--cutonyield <>", <> specifies the # of sigmas of the cut 
	(default 10).
	"--minimumqrts <>", <> specifies the minimum good quartets 
	(default 1000)
	Also modified the replayflags.in to reflect these.
     *  Split FillAnalysisTable() into CheckIdenticalAnalysisRecords() and 
	FillAnalysisTable(). The former will be called at the beginning of 
	the replay, such that it quits if an identical previous replay is 
	found. The latter is called at the end of the replay, right before 
	all the results get filled into the DB. So if the replay crashed 
	in the middle (from time to time on the batch farm), no corrupted 
	records get filled into the analysis table.
     *  When retrieving the slope values from pass "A", do a NULL test on the 
	returning value. If NULL set the slope to zero.
     *  Kaz put in another fix in the slope correction routine. The 
	NA detector "goodness" flag was applied on the French detectors 
	before, which caused some bogus slopes in pass "B" for French octants.
     *  Some new histograms were added to get ready for the next run.
	a) offline:
	   LTD error and DMCH alerts (4+4 1d)
	   ave/diff of theta angle in x and y(2+2 1d)
	   total yield for NA and French rings (16*2 1d)
	   yield/asym of lumis (8+8 1d)
	   Pivot halo rates (1 1d)
	   ADC vs ring/tube# for individual octs(8 2d)
	   CFD TDC hits vs ring/tube# for individual octs (8 1d)
	   MT TDC hits vs ring/plane# for individual octs (8 1d)
	b) realtime:
	   LTD error and DMCH alerts (4+4 1d)
	   lumi asymmetry (8 1d)
	The default histogram binning file g0hists.in has been modified.
	Rescaled all the detector asymmetry histogram in the units of ppm,
	instead of "%". Left the target boiling histogram unchanged.
     *  Fixes at various places in the code to stick to a single convention of
	the variables:
	a) yield: qrtsum_yield/4.
	b) diff: qrtdiff/2.
	From now on all the histograms, ntuples and DB will be according 
	to this convention.
     *  Added the raw 30Hz event counter printout at the end of the replay.
     *  As Silviu requested, we now also impose 10 sigma yield cut on the NA 
	"direct" scaler data and the SDMCH data for target boiling studies. 
	Only the asymmetries and yield of the "good" quartets are histogrammed.
     *  Added in Retief's code of the VME TDC (laser start, beam pickoff, 
	etc ..) analysis. Updated the TS channel map in beam.in.01-23-03. 
     *  Implemented Damon's trick to retrieve enum variables from the DB. 
	Added a column in the database look up table to specify whether the 
	column is enum or not. If yes the SQL translator will put "+0" in the 
	"select" query, such that atof() is possible on the returned string.
     *  Fixed a bug in constructing the realtime histogram.
     *  Encapsulated the Translate(), Query() and Process() methods into some 
	overloaded Process() methods.



G0Analysis Release 0.23
     Paul King; 2003/12/18, 10:00

     This is a late tag.  This version represents the state of the
     analysis software as of 10 September 2003.

     Changes since release 0.22:
     ===========================
     *  A bug was reported by Silviu, that there were some very short runs 
	that sneaked into the beam table, with non-null values. It looked like 
	the n_quartet>1000 cuts were applied on the FPD and slopes data, before 
	filling the DB, but they weren't applied on the beam data. Fixed.
     *  Added the TFrasier classes to the repository
     *  Added some realtime GUIs into the repository.
	      G0RateMonitor (updated version)
	      G0RealTimeMonitor
     *  Fixed a bug in G0SQL::IsEnum(). If it fails to find the matched variable 
	from the predefined variable list, return false (assuming not enum).
     *  Moved color_thresh.in (the color code for the rate monitor) to 
	Analysis/prminput/. Changed the path in G0RateMonitor.C.
     *  Junho added a bunch of "polarized source" parameters to be read by CODA,
	and subsequently analyzed by g0analysis.  These parameters include IA,PZT 
	voltages and injector BPM values.  Added pivot halo in/out status and
	position feedback on/off status to be read also.  This was tested with 
	a mock G0DAQ run, and were seen to be working.  
     *  Added a new global variable, "kInG0BatchMode", which is true when there
	is no DISPLAY environment variable.  This variable is used to exit the
	analyzer when user input is needed but the code is executing in a
	non-interactive environment.
	The only two places in which this is used are in setting the French
	configuration by hand, and in entering the password for the MySQL database.
	Also, a new method has been added to G0FrenchDetector; the method,
	SetFrenchCfgWdFromFile, will be used by the real-time analyzer to get the
	French configuration directly from the DAQ configuration file.
     *  The G0RealTime analyzer has been updated to include the EPICS and control
	event processing code, and several minor improvements from the G0Analysis
	analyzer.
	Also, the G0RealTime analyzer will not time-out when waiting for the ET
	system to transfer events.  However, if the ET system is not running when
	the analyzer is launched, the analyzer will exit immediately.
     *  The method used to decode the French configuration buffer was modified to
	use the length of the buffer instead of the run number to decide which
	decoding pattern to use.
     *  An error in copying the French configuration information from the CODA
	event into the internal buffer was fixed.  This error caused data words
	beyond the end of the French configuration to be copied into the buffer.
	However, this copying was done in a "safe" way, so the data was not corrupted
	by the error (there was just extra unused data in the buffer).
     *  Fixed an error in the FPD "group by" tranlation syntax. 
	So now "[y,a]_det#[pid](err)" will calculate the weighted average and errors 
	on the server for FPD ring # (grouped by detectors, summed over octants) 
	and retrieve them as two columns.
     *  Fixed a bug related to translation of the self-join of the slope table.
     *  Just commented out a debugging output in TFrasier.
     *  In G0SQLResult, implemented "GetArray()" by the column index. 
        Before we only had GetArray()  by column name, and the code got confused 
	when two columns had the same name,  e.g. "m.value".
     *  In G0SQL, in processing the cuts string in the SQL query translator, now also
	take  "+-*/" as the separators between variables. This is to allow the 
	arithmetic  operations in the cuts. Also implemented a literal translation in 
	cut string enclosed by "##". Especially since "." has been reserved to separate 
	table alias and the field name, so use #5.98# if one wants to use the 
	decimal point in the cut string.
     *  Changed the default value of kInG0BatchMode; now all versions of the
	analyzer (and ROOT scripts) will default to kInG0BatchMode==kFALSE
	(in other words, the program is not in batch mode).
     *  There was an incompatablility in the 120 Hz analysis mode resulting from
	a change in the definition of the G0Cuts object.  The internal copies of
	the G0Cuts object are now created with the same arguements as the original.

	The GetReplayPrms functions were changed so that "--cutonyield off" would
	disable the "10-sigma" cut on the detector yields.
     *  Added a variation of the G0Error function which allows the exit status
	to be changed.  Also created an enumerated list of G0 error codes for
	the 120Hz analysis.

	The analyzer will now exit with status "10" if the data file was taken
	in 120Hz mode and the analyzer is not configured for 120 Hz mode.  The
	analyzer will exit with status "11" if the analyzer is configured for
	120 Hz analysis but the data file was not taken in 120 Hz mode.
     *  Some modification of the Fraser class.
	1) To avoid memory leak when creating the histograms with the same name, in
	TFraserHist::PreparePlot(), create TH1 with the name of the input string arg. 
	It should be transparent to users.

	2) Minor fixes in TFraserGraph. Only create the TGraph obj of in/out 
	ihwp if the number of points is >=1. This fixes the plot with only one 
	ihwp state.
     *  In G0Hists.h, now redefine all histogram pointers as the base pointer TH1 to 
	ease the coding. Fixed related hooks in the GUI applications.
     *  Some modification of the realtime and offline engine.
	1) In the realtime code, now by default it loads all the calibration data from 
	the database. Define a kLastRun (999999), by which the realtime code searches 
	for the most recent calibration set. Some minor modifications in 
	G0SQL::SelectCalibrationID(). Note for later maintaining the calibration 
	tables, we need to make sure that the most recent set in each sub table 
	last_run_id = run_id(kLastRun). Minor modifications in inputfile_lookup.in to 
	reflect this convention.

	2) Modified the database access flag "-d". Now -d takes argument as the 
	DB access level <off/r/rw>. Modified the default flag file replayflags.in so 
	the default access level is "readonly" -- retrieving the calibrations, cuts. 
	Only "rw" level writes to the database. Modified the related hooks. If you 
	want to run replay engine without accessing the DB, be sure to use "-d off".

	3) Added a test in G0RealTime on 120Hz_Enabled. Skip all 120Hz events
	processing in realtime engine.

	4) Added some asymmetry correlation histogram into the offline_hists list, 
	including A_fpd(na/fr) vs x, A_fpd(na/fr) vs y, A_fpd(na/fr) vs q. Modified 
	the histogram definition file g0hists.in. Modified the format to include 
	the x,y axis title/unit when creating the histograms.

	5) Redefine all histogram pointers in G0Hists class as the base pointer TH1 to 
	ease the coding. Fixed related hooks.

	6) Update SumDet_Raw in NA and French Detector classes to use the PID weight 
	when computing rates for the rate monitor. Insert CalcPIDSums()
	in G0RealTime before computing the rates. Removed the obsolete CalcRates() 
	in the Helicity class.

	7) Added the quartet-averaged x,y and charge into the quartet ntuple.

	8) In G0BeamMonitor assignment operator, now also assigns the MPS, 
	QRT and QRT_phase counters.
     *  Added g0realtimemonitor, g0ratemonitor and makefraserplots into the 
	default compilation target list (along with the existing 
	g0analysis, g0realtime, g0acm, g0root). So now TFraser classes should 
	be compiled into libG0.so by default.

	Test compiled on both solaris and linux farm machines.
     *   Some realtime code work:
	1) Added the display of Lumi asymmetry histograms into realtimemonitor.
	Also in the stripchart panel, added 3 lumi scaler vs time plots.

	2) Added two 2d histogram (octant:detector) for LTD errors and DMCH 
	alerts respective. Put them on the "Misc" panel.

	3) Added a launcher button on the realtimemonitor control frame for the 
	ratemonitor. Put a kludge test on whether the process is running and 
	launches the g0ratemonitor GUI if not.

	--- An important code fix for the upgrade to ROOT3.05/07-gcc3.2: 
	It was discovered that the anode current monitor dictionary compilation 
	failed with the new ROOT and gcc3.2. It turns out that a constant defined 
	in the linear regression class, "dim", shadowed a variable of the same name in 
	a ROOT include file DataMbr.h. So instead defined a static constant kDimension
	as the internal member of G0LinearRegression class.

	--- Misc modifications
	1) Minor changes in G0FrenchDetector.C, G0ROOT.C to remove the warning 
	message during compilation.

	2) Modified the fastbus ntuple builder flag to --fastbusntuple <all/na/fr>

	3) Added 4 1D histogram into offline_hists --- total FPD proton yields and 
	asymmetries (na/fr).
     *  Minor changes in G0FrenchDetector.C, G0ROOT.C to remove the warning 
	message during compilation.
     *  Some realtime code work:
	1) Added the display of Lumi asymmetry histograms into realtimemonitor.
	Also in the stripchart panel, added 3 lumi scaler vs time plots.

	2) Added two 2d histogram (octant:detector) for LTD errors and DMCH 
	alerts respective. Put them on the "Misc" panel.

	3) Added a launcher button on the realtimemonitor control frame for the 
	ratemonitor. Put a kludge test on whether the process is running and 
	launches the g0ratemonitor GUI if not.


G0Analysis Release 0.24
     Paul King; 2003/12/18, 10:00

     This is a late tag.  This version represents the state of the
     analysis software as of 25 October 2003.

     Changes since release 0.23:
     ===========================
     *  Removed the explicit dependence on the run number from the functions
	G0FrenchDetector::Get_French_DACs, G0SQL::SelectCalibrationID, and
	GetInputData.
     *  Added a function G0ReplayPrms::GetThisRun, which returns the current
	run number.
     *  Modifed the G0120Hz class so that all phases of the 120Hz object use
	the same G0ReplayPrms object.
     *  The function G0FrenchDetector::Get_French_DACs now uses a test on the
	number of data words to determine which decoding scheme to use.
     *  The function G0SQL::SelectCalibrationID now uses G0ReplayPrms::GetThisRun
	to get the run number for the database lookup.
     *  Updated the calls to G0FrenchDetector::Get_French_DACs and GetInputData
	in G0Analysis.
     *  Added code to do the decoding of the NA monitor detectors.  Probably
	there will be additional work inside the G0NAMonitor class to handle
	the "unused" time bins properly.
     *  Tried to clean up all the unnecessary "#define" statements. If needed 
	replace them by internal static constants. Removed the 
	redundant constants in variaous classes. 

	Defined all global constants used as array sizes as static constants 
	(instead of external) in G0Globals.h, and declared the values.
     *  Fixed some mistakes in G0NADetector.C where kNAOctants(4) were mistaken by 
	kOctants(8).
     *  Fix a bug in G0FB_deadtime::ReadNAFitParameters(). The original character
	array (holder of the filename) was too small to store a long path and caused 
	segv. Replaced the array by TString instead. 
     *  Added HallA IA/PZT, prebuncher, laser attenuator, Master slit position, machine 
	energy loss, raster dimensions, HallB TACO, and PITA voltages to EPICS 
	dictionary.
     *  Modified the loading of DNLs for both electronics (now from na_dnl_data 
	and fr_dnl_data tables).
     *  Defined fCoilModActive and fHVInterlockTripped variables and their getters 
	and setters in G0TriggerSupervisor.
     *  Take gzero-eng-4threplay as the default DB.
     *  Sarah and Damon discovered a bug in WriteToBeamTable that the 
	measurement_type_id for lumi asymmetry was wrong ("a" instead of "al"). 
	Fixed.
     *  Added Jeff's na_dnl calibration file into prminput/. Note this file is NOT 
	read by the replay engine. I stored the file here for later loading by 
	the database creation script.
     *  Added NA buddies histograms in G0Hists
	Also added a routine that dumps the NA monitor channels on the screen (this
	is not complete and will change in the future)
     *  Added new memberfunction CalculateNAdt_factor70 in G0FB_deatime to allow to run 
	with either the 70ns or the "regular" singles correction by using the replay 
	options. Note: Before it was necessary to edit and recompile the code in order 
	to switch between corrections. 
	Also fixed a small error in the rates calculation (the effect of this mistake 
	was very very small).
     *  Added the filling of tartan table histograms (in offline_hists directory) for 
	the french electronics
     *  Added some double cast in G0Hists::FillNABuddiesFinalHists() when 
	one takes sqrt(NumQRTs). This solved the compilation error with gcc3.2.



G0Analysis Release 0.25
     Paul King; 2003/12/18, 10:00

     This is a late tag.  This version represents the state of the
     analysis software as of 22 November 2003.

     Changes since release 0.24:
     ===========================
     *  This submission imports some structrual modifications Kaz and I made in 
	the linear regression code.

	All the c routines from the numerical recipe to solve a NxN matrix 
	equation were taken into Matrix.{C,h}. A linear regression 
	class has been coded (LinearRegression.{C,h}) to perform the regression 
	on one variable. Then the original G0LinearRegression class has been 
	changed into a wrapper of [8][16] LinearRegression objects, (and will late 
	be expanded to lumi detectors as well).

	This change will allow flexibily for future modifications. Hopefully the 
	code is more robust and readable.

	The calculations were cross-checked against the older version of the code.
	In doing so a few (minor) bugs were discovered and fixed.

	1) When setting the "goodness" flag in the quartet cut 
	(G0NA/FrenchDetectors::SetGoodQRTFlg()), all the 4 MPSs are 
	supposed to be tagged as bad when calculating slopes, if any of the 
	MPS fails. I forgot to do that, so some of the yield slopes have 
	been calculated including some "broken" quartets. In principle this 
	should be OK, but it is not 100% consistent with the asymmetry calculation. 

	2) It was discovered there exist small numerical difference in
	slopes between the old and the new code. Eventually it was traced down to a
	double to float then double casting in the old regression code (xvar[] was
	defined as float ...). Fixing that reconciled the differences. 
     *  Initialize all the TGraph pointers to NULL in the c-tors for safety.
     *  Added a new histogram-class into G0Hists (ConstructNAMHists and FillNAMHists) 
	to fill the readout of the newly installed NA Monitor LTDs into histograms. 
	Two new flags were added to g0analysis (--NAMhists and --DumpNAM). The first 
	creates the histograms, the second dumps the NAM readout to stdout (added by 
	request of Brian). The two can be run together, but also work independently.
     *  Changed the carbon target type "c" to tungsten radiator "r".
     *  Changed DB table for EPICS g0_pivot_halo to 'target'.
	Changed EPICS SMWPOSC to SMRPOSC (Hall C chopper slit readback).
	Added following EPICS variables:  psub_aa_pos, psub_ab_pos, itov1out,
	itov2out, itov5out, itov7out, IGL1I00PStrans, SMRPOSA, SMRPOSB.
	Changed 'polarized_source' to 'polarized_source_measurements' in
	AddEPICSTag for many of the polarized source EPICS variables.
	Note:  Still need to add specialty code to deal with additional single-valued 
	EPICS variables being written to 'polarized_source', e.g. HC:P_ONOFF.
     *  Changed EHCFR_LIXWidth & EHCFR_LIYWidth storage tables to
	'target_measurements'.
     *  Added this derived class from TGraphErrors. It takes a G0SQLResult object 
	in the constructor and makes it a TGraphErrors object, so it will be 
	transparent for later processing, eg. making a multigraph. Modified from 
	Julie's grepdata::CreateTGraph() method. Added a simple example in 
	MakeFraserPlots. Note it is different from fraser graph, where the results are
	IHWP in/out sorted.
     *  Some minor fixes in the graphing classes. redefined some global constants as 
	the static constant member of TFraser. Generate dictionaries for 
	TG0SQLResultGraph.
     *  Target update from Lars.  Added flyswatter position and g0target pumpspeed.  
	In epics.dictionary, added above 2 as well as flyswatter and radiator 
	temperatures.
     *  1. Made some modifications related to the coil modulation analysis. Since 
	there is a trigger bit to indicate a coil pulsing MPS, we have decided to 
	perform 2 replays for each run, one analyzes the coil pulsing events only, 
	and the other analyzes only the NON-coil pulsing events.

	a) Set "CoilMod" bit and "HVTripped" bit in 
	G0TriggerSupervisor::ProcessEvent().

	b) Added "--coilmod" replay flag to switch on the special analysis mode in 
	which only the the coil pulsing events would be analyzed. (By default this 
	flag is off.)

	c) After processing the TS buffer, if "--coilmod" is on but the trigger bit 
	is off, OR, if "--coilmod" is off but the trigger bit is on, skip the event.

	d) In G0SQL, made the changes related to the additional "beam_mode" flag in 
	the analysis table. 

	e) The analysis table is filled with the additional beam_mode flag. The 
	physics table fills are different for two cases. When "--coilmod" is on, 
	we only fill the slopes and beam tables.

	2. Put the test on HVTripped bit in PassEventCuts()

	3. Implemented the filling of fr_cfd_rates and fr_mt_rates tables at the 
	end of the replay. The rates are based on the SDMCH data, with "cfd" being the 
	CFD_total-MT_raw and "mt" being the MT_raw (same convention as used in the 
	offline histograms).
     *  Modified a few graphing style options.
     *  To set the Y scales of the graph by hand, enabled SetMaximum() and 
	SetMinimum() functions. Note the default scales has been set in the object
	constructor. So one only needs to call these setters if want to change the 
	scales. It can be called either before or after the Draw() function.

	Fixed a bug in the TSuperFraserGraph c-tor: followed the changes made in the 
	TFraserGraph c-tors, moved PrepareGraph() from the Draw() method into the 
	object c-tors. Otherwise it would cause segv when Draw() is invoked.
     *  Added a supporting class TIHWPInOut to facilitating the graphing of physics
	asymmetry with IHWP In-Out/In+Out. Compiled into libG0.so
     *  1) Added x,y,thx,thy histogram for coil pulsing events only. Also added
	Aq vs position differences histograms. Modified the histogram binning file
	g0hists.in.

	2) Added "--online" flag. In this mode, the code will process all (both 
	coilmod and non-coilmod) events. Thus we only need to run the "online"
	engine once.
     *  1) Fixed the database read access problem for the realtime engine.

	2) Now by default we load the french single deadtime correction parameters 
	from the database. (fr_deadtime_data table wasn't avaiable in
	gzero-eng-4threplay db).

	3) Take wmg0://gzero_comm as the default database.

	4) Added Junho's beam feedback code into G0Analysis CVS. Modified the Makefile 
	to compile g0beamfeedback if EPICS library is available. Note this is the 
	"forever" version of the code, in which the position and charge feedback 
	are computed separately.

	5) Fixed the problem when resetting lumi stripcharts in realtime monitor.
     *  Modified the main code for the realtime engine, such that as soon as 
	g0realtime gets started, it will open and write to the TMapfile. Thus the 
	engine will take control of the mapfile, all clients will only have the 
	read privilege.
	Removed the redundant varaible firstevent and the test on it.
     *  Forgot to Update the TMapfile after all histograms get created.
     *  Removed the historical "g0monitor" from killall().
     *  We decided to remove the beam feedback code from the G0Analysis repository. 
	The setup of the Makefile inadvertablly compiles libG0.so with the EPICS 
	library on g0ana, so all user codes that compile against libG0.so do not 
	work without the linking to EPICS library. It's just too much effort to 
	alter all the user Makefiles ...

	We will create a separate directory structure in CVS for the feedback code 
	and with a standalone Makefile for it.

	Reverted the G0Analysis Makefile to its original state.
     *  Add a default value for the "executable type" variable in G0ReplayPrms.
	This variable is used internally to distinguish between realtime and
	online/offline analysis processes.




G0Analysis Release 0.26
     Paul King; 2003/12/18, 10:00

     This is a late tag.  This version represents the state of the
     analysis software as of 1 December 2003.

     Changes since release 0.25:
     ===========================
     *  1. Put in the new beam scaler map beam.in.11-21-03. Modified its format
	   such that the calibration values is no longer present in this map. Also 
	   modified the scaler map lookup file to reflect this change.
	2. Moved the assignment of monitor counters, e.g. NUM_BPM, into the 
	   scaler mapping routine MapIt_beam(). (Before I had them in the calibration 
	   parameter loading routine.) This is to avoid the potential mistake, say if 
	   one of the monitors's calibration is missing in the DB. As a consequence, 
	   reversed the order of MapScalerChan() and the TMapfile construction 
	   in the realtime code.
	3. Modified the routines related to the halo and lumi histogramming. Replaced 
	   the historical "h_pvhalo_rate" in the offline histogram by "h_halo_rate",
	   and just fill Halo1 rates for now.
	4. Fixed a bug in FillRealTimeHists(), where TH1::Clear() was used instead of 
	   Reset(), which caused the subtle segv of the realtime engine.
	5. Modified the realtime display.
	   1) Changed the old charge asymmetry panel into a "parity quality" panel, 
	   which displays the histograms of charge asymmetry as well as the beam 
	   position differences in x and y projected onto the target.

	   2) Modified the beam stripchart panel. Removed the somewhat redundant 
	   x_g0, y_g0, x_g0b, y_g0b and replayed them by x_targ and y_targ. Added 3 
	   pads for halo monitors, with halo1&2, halo3&4, halo5-8 overlaid. Added 3 
	   pads for lumi monitors, with lumi1-4, lumi5&6, lumi7&8 overlaid.

	   3) In the ratemonitor stripchart, for some reason I SetTimeDisplay() before 
	   Resetting the gStyle (and somehow it was able to run in the older ROOT).  
	   The new ROOT helped me to find this bug!

	6. Minor changes in G0Globals (declaring the VME_TDC related counters 
	   constants).
     *  Changed the way the beam calibration parameters were handled. The labels 
	of each monitors is now loaded when reading the scaler map files MapIt_beam(). 
	Later when accessing the DB, we will do string matching of the monitor 
	names and then assign the calibrations constants. (Before the monitor 
	names were also loaded from the DB --- LoadBeamCalibPrms()).
     *  Modifications in the LoadBeamCalibPrms() in the string handling.
     *  Split the original GetInputData() into two steps: 
	GetInputData() and then LoadCalibrations(). Now the function calls in the 
	realtime and offline engines are:
	GetInputData() ---- get the text file input
	MapScalerChan() --- establish the electronics map, defining the monitor 
			counters, etc
	LoadCalibrations() --- read from the DB for the calibrations
     *  Junho gave me the updated scaler 1 map for the halos. I updated 
	beam.in.11-21-03.
     *  A few minor changes related to the filling of the beam table in the DB.
	Forgot to remove the Chop() function from LoadMonitorTable (no longer needed
	since we've fixed the monitors table). Also changed the halo yield 
	measurement type to 'yhp' in WriteToBeamTable, since now all the 
	halo monitors are counting.
     *  Changed the updated scaler map range starting from run 17017.  
     *  Found a silly mistake when loading the beam monitor calibrations --- 
	it broke from the while loop after the clk calibration got read in. Fixed.
     *  Added G0 laser phase jitter to monitored EPICS variables.
     *  Changed the binning of charge asymmetry and position difference histograms 
	to do better in the RHWP scan and IA scan.


G0Analysis Release 0.27
     Paul King; 2003/12/18, 10:00

     This is a late tag.  This version represents the state of the
     analysis software as of 6 December 2003.

     Changes since release 0.26:
     ===========================
     *  Added the members fCustomFileStem and fCustomFileStemFlag + getter and setter
	to the class G0ReplayPrms in order to be able to give rootfiles names other
	than the standard names.

	By using the new flag "--outputstem MyFileStem", one will override the standard
	name chosen by g0analysis. The file will be named MyFileStem.####.root and
	will be found as usual under $G0_ROOTFILES_DIR.

	This will allow more flexibility and hopefully prevent the shift worker from
	overwriting by mystake files generated by the analysis team.
     *  Tried to fix the memory leak of the realtime display. Found a few 
	places that I incorrectly delete the dynamically loaded pointers. 
	Did more careful clean up after each Draw(). Now at least the ROOT object 
	table is happy. But there's still memory leak and I traced it to  
	"histogram Draw() + TCanvas Update()". Haven't found a fix yet.
     *  I made some modifications to solve the realtime display memory leak problem
	related to TH1::Draw() + TPad::Update() for ROOT 3.05/07. Using the ROOT 
	"insider" tricks to get around. basically
	create h1
	h1->Draw
	loop{
	  h1->fill
	  c1->Modified
	  c1->Update
	}
	This avoids Draw in the loop, but the graph still gets updated. 
	See comments in FetchHistogramFromMapFile and DisplayHisto/DisplayAsymHisto.
     *  Now do NOT impose the 10 sigma cuts on the 1st NA ToF bin. Occassionally 
	the latch got stuck (not disatrous according to Brian), which produces 
	an empty 1st bin. The 10 sigma cut we had before would tag this whole 
	ToF as bad, which is not desirable now.
     *  Defined a variable "brian" to switch between NA/French to en/disable the 
	10 sigma cuts on the 1st ToF bin.
     *  Tried to remove the histograms before the while loops, seemed to 
	fix the segvs in the realtime monitor.
	Before I had those at the end of the while loop, but 
	somehow the ROOT system break skipped those ...
     *  Changed all multiplicative factors for NA ADC channels from .1 to .0534
	This is the new conversion factor to convert NA ADC bins to microamp in g0acm



G0Analysis Release 0.28
     Paul King; 2003/12/18, 10:03

     Several bug fixes found during the first part of the engineering run.  Also,
     this is the first introduction of the G0FrAmDetector class for the hybrid
     electronics.

     Changes since release 0.27:
     ===========================
     *	Modifed Math_Func.C and WriteToSQLDB.C to add extra protection to keep
	from writing "NaN" (not a number) into the database.  There seem to be
	some divide by zero errors which had not been caught; perhaps this is
	due to the new code which allowed events with zero in one or more timebins.
     *  Modifications to the G0FrenchDetector class were made to support the addition
	of the new G0FrAmDetector class for the French-American hybrid electronics
	instrumenting detectors 14 and 15 of octants 1, 3, 5, and 7.
	The G0FrAmDetector class is derived from the G0FrenchDetector class.

	The principle changes are to the mechanisms used to loop over the detectors
	and octants within the French detector class.
	Several static variables of the G0FrenchDetector class were made into global
	variables.
     *	Added "_120Hz.in" to the prminput directory; this file is needed by the 
	120Hz analyzer. It sets overriding flags for the 120Hz analysis. For the time 
	being, the version to be used on the analysis machines should not be the 
	default one. Yet, the environment variables point to the prminput standard 
	directory, and therefore, this file must exist.
	Its impact on the normal analysis is null.
     *  Modified G0BeamMonitor::CutTripEvent(). The original algorithm counted
	2000 MPS after the beam recovery based on the hardware MPS counter, which
	had the side effect in realtime processes -- MPS counter got reset when a 
	new run gets started. So now uses only the software counter beamtrip_counter.
     *	Fixed a typo in the cut function in G0BeamMonitor class.
     *  Initialize the beamtrip_counter in the c-tor to NumEvSkipAfterTrip so that
	at the beginning of the program the first 2000 MPS don't get cut.
     *	Added helicity correlated differences in projected beam positions on target,
	D_x_targ, D_y_targ, beam angles D_th_x, D_th_y into the quartet ntuple to 
	ease Kaz's HC evolution analysis.
     *  Added Brian's technique into G0NADetector::ProcessEvent(). Always take 
	the sum of the first 2 bins and divide by 2, to avoid the zero counts in 
	the 1st bin.
     *	Removed some debugging printouts in the routine CutsOnDetectors
     *  For consistency, in the overloaded method ProcessEvent(), called by the 
	realtime engine, also uses Brian's trick to remove potential zero weight 
	bins.
     *	In ResetCounters() routine (called at the beginning of each offline replay),
	initialize the beamtrip_counters according to the values set in the c-tor.
	Two days ago I modified the beam trip cut routine to fix the realtime 
	programs, but apparently broke the offline engine, due to the inconsistent 
	variable initializations. This bug caused the first 2000 MPS of a run to be
	tagged as bad events, so for very short runs one might find 0 good 
	30Hz events and quartets ... 
     *  Commented OpenRateFile(), CloseRateFile() calls in G0Analysis.C and 
	G0Realtime.C. The opening and closing of the rate output file will happen 
	in UpdateRateFile() routine, if one has accumulated enough events for one
	iteration. This is to avoid creating empty file, if the DAQ is running with 
	non-30Hz mode.

	Moved the function call of UpdateRateFile() outside the beam current cut in 
	the realtime engine, to enable the rate display for low beam current runs.
     *	In realtime engine, also moved the PID sum calculation out of the
	beam current cuts.
     *  For the CFD and MT hits histogram, removed the overflow and underflow 
	cuts. This is expected to fix the discrepancy in rates displayed in 
	the realtime monitor. 
     *	In the main routine, the NA deadtime correction parameters, as well as some 
	fastbus paraemters, were not loaded in the initialization!!! This happened 
	when I splitted GetInputData into three different routines earlier 
	this month. Only NA deadtime correction is affected
     *  Modifed the TDC histograms for the realtime display to put a cut at zero
	and a cut at 4090.  This is to correct the realtime CFD and MT hit displays
	to onlyt show the number of valid hits in each channel.


G0Analysis Release 0.29
     Paul King; 2004/01/13, 14:30

     Code modifications before the beginning of the January running period.
     Major changes are the inclusion of analysis code for the Franco-American 
     hybrid electronics (DMCH9), and several fixes for deadtime correction.


     Changes since release 0.28:
     ===========================
     *  Added the test on the event type before filling different types of realtime 
	histograms (eg. 30Hz, fastbus, etc).
     *  Fixed a problem in the NA error counter realtime histogram. Now the 
	error counts should show up in the realtime (group "MISC").
     *  Added the NA scaler input file "nascalersMap.in.12-19-03.tgt_boiling"; this
	input file has the decoding for direct meantimer scaling for NA meantimers
	1, 5, 2, and 6 in monitor scalers 1 and 2 (the 16th and 17th scalers in ROC4).
     *  Updated the DB with the new NA deadtime single rates parameters filled in.
	Now switch the default NA deadtime correction flag to "dog_singles".
     *  Fixed a bug related to NA deadtime correction. When selecting the dog_single 
	option, the singles correction was overwriting the dog correction.  Now both 
	correction are applied.
	Also changed a cut in G0FB_deadtime.C applied to FB data when filling tartan 
	bins histograms. The cut now selects a region of 96ns instead of 93ns as in 
	the previous run and it is also shifted in order to have the cut far from 
	the proton peak.
     *  Modified the way the calibration_id is fetched. Since the calibration and 
	run table have become large, the original "self-join run table 14 times" trick
	take longer to load. Split the query into 7 separate ones. This helped to 
	speed up the initialization.
     *  Fixed a unit conversion problem (%->ppm) in one of the NA proton 
	asymmetry histogram filling.
     *  Updated the routines to decode and process the FrancoAmerican electronics
	data.  Several of the G0FrAmDetector methods have been redefined instead
	of taking the default behavior from teh G0FrenchDector class.

	The other classes are not yet aware of the G0FrAmDetector class, except
	for the G0Hists.  The ToF histograms are generated, although the DNL
	corrections are not applied to the ToF histograms.  No other analysis
	is done on the FrAm data.
	The FrAm ToF histograms have the same layout and naming as for the French
	detectors.
     *  I forgot to add the halo calculation into the assignment, + and - operators
	in G0BeamMonitor class. All halo rates in the DB are zeros before this. Fixed.
     *  Fixed an array index mistake in the linear regression code. This only affects
	the corrected yield calculation and we haven't used the routine so far.
     *  Added the lumi slope calculation and correction to the code. Implemented some 
	temporary hooks to write the values to a file.
     *  Added a new TS data structure file to deal with the change of subbank 
	size for the Yo TDC. Tagged all additional data words as "spared".
     *  Modifed the ROC decoding routines to avoid a double incrementing of the
	word counter for the ROC3 bank when operated in the "split bank" mode
	(as when running in the ts_full2 or TSdmch_FrAm run types).
	This should fix the problems analyzing those run types (the problem was that
	the ROC4 data was not unpacked correctly).
     *  A few fixes to enable doing the 2nd pass replay of the slope correction.
	When checking the flags in the analysis table, loosen the requirement a bit --
	instead of checking the full path name of input files, check the "date" part
	only.
	Modified the Drawing option of the fraser plot a bit to reduce the size of the
	data marker.
     *  Fixed a string composition problem in G0SQL::CheckIdenticalAnalysisRecord().
     *  Found a mistake that caused the halo asymmetry not being calculated 
	correctly and thus not appear in the DB.
	As a starter, Halo rates are not normalized by the beam charge. The raw
	rates and the raw asymmetries will be filled into the DB.
     *  Normalize the halo rates by the beam current before computing asymmetries. 
     *  Fixed a dumb path problem in the lumi slope output file.
     *  Put a small fix from Angela in combining NA tartan rates and folding with 
	the deadtime. If Front coin but B single, then multiple by 48 ns deadtime, 
	instead of 32.

G0Analysis Release 0.30
     Paul King; 2004/01/18, 16:30

     This is a short set of changes; the principle changes are to the realtime
     analysis, where normalized values are now displayed in some displays.
     Also, there was the addition of a few EPICS variables for beam current in
     the other halls, and for the G0 electronics cage temperatures.
     Some additional clock signals were added to allow study of the strange
     variation in the 4 MHz V/F reference clock signal.

     Changes since release 0.29:
     ===========================
     *  Added several EPICS variables to the G0EPICSEvent decoding; these are
	for: Hall A MPS BCM beam current, Hall A BCM beam current, 
	Hall B 2C24A beam current, Hall B fcup nA beam current,
	Hall C MPS BCM beam current.
     *  Added the decoding of the additional V/F clock signals. Changed the 
	data memeber G0BeamMonitor::clk into an array. Also added them into the 
	beam ntuple. To keep things consistent we are still filling the value of 
	the old clock (V/F5) to the database.
     *  Added a new beam scaler map file with the updated clock channels. 
	Changed the format of the "clk" channels. Modify the decoding of the map 
	file. Modified the lookup file as well.
     *  Fixed a bug related to the --cutonyield flag. This bug caused the code to 
	crash on values other than "off" and "10".
     *  Changes requested by Silviu and Herbert. Now compute the normalized FPD 
	rates and lumi yield and put them into the realtime displays. So now the 
	FPD rates shown in the STRIPCHART panel in the ratemonitors are in kHz/uA.
	The lumi yields in the "beam stripchart" panel in the realtime monitors are 
	also V/F bits/uA.
     *  Fixed a bug related to the NA error counts realtime display. They should 
	now show up properly in the "MISC" panel.
     *  Added three EPICS variables to the decoding list in G0EPICSEvent; these
	are the variables for the G0 electronics cage temperatures.


G0Analysis Release 0.31
     Paul King; 2004/01/31, 12:30

     The major change in this version is to add a new ET client mode
     which allows filtering of events based on the event type before they
     go into the decoding routines.  This should allow for faster processing
     of MPS events in the beamfeedback analysis.
     The beam input files (used for channel assignment) were found to
     have changed in the syntax used to describe the CLK channels; this
     change made the analyzer crash when analyzing old data files.  The
     old beam input files have been updated to use the new syntax.
     There were serveral other minor changes and improvements in the
     realtime and offline displays.

     Changes since release 0.30:
     ===========================
     *  Added the Halo target position to the EPICS events and the EPICS 
	decoding.
     *  Modified G0BeamMonitor, the scaler map files, etc, to handle the 
	decoding and computation of the additional word for Hall A BCM.
	Modified the map lookup file.  Added decoding and some computing 
	routines in G0BeamMonitors.
     *  Implemented Paul's ET "filter" client mode in the TEtClientBuffer, 
	such that we can have a thinner client (e.g. g0beamfeedback), that 
	only reads the type 1 physics event. Corresponding changes in the 
	feedback code as well.
     *  Minor changes in the realtime displays (cuts in beam current).
     *  Minor fix in graphing style for IHWP in/out plotting in the GUI 
	classes.
     *  I have modified the "CLK" line in the beam input files for the 
	2002/2003 data files to match the syntax used in the current beam 
	input files.  The change in syntax was causing the analyzer to 
	crash when the older data files were analyzed.


G0Analysis Release 0.32
     Paul King; 2004/02/05, 16:25

     The main change for this release is the inclusion of the new
     decoding and analysis routines for the YO VME TDC data.
     There are also two additions to the EPICS event decoding:
     hel_mag_status, which looks at the integrated helicity bit 
     difference over time, and the voltage readback from the dummy
     helicity Pockels' cell.


     Changes since release 0.31:
     ===========================
     *  Added a new value that looks at the integrated helicity bit 
	difference over time.  EPICS variable: hel_mag_status.
	The value takes h+ - h-  which can be either 0 or 2, and sums
	these numbers up over time.  In principle, the integrated value
	should be 0.  If it is not 0, that indicates that there were 
	"broken quartets".
     *  Modifications to TIHWPInOut in the GUI:
	1.  Added a few min/max getters and setters, to allow user-defined
	    Y axis scaling.
	2.  Added options in Draw() to allow one to overlay In+Out, 
	    In-Out graphs.
	3.  Added the getters for individual values, such that the 
	    asymmetries can  be output into text form.
     *  Switched the halo offline histogram to display halo3 rates.
     *  Added the voltage readback of the dummy helicity Pockels' cell
	to the EPICS events and decoding.  
	This can be used to see the effect of electronic pickup of a 
	helicity signal that is shifted in time with respect to our 
	normal quartet helicity signal.
     *  Submitted Klaus' changes to the VME TDC decoding and analysis 
	routines to work with the extra TDC channels for the six YO 
	signals: BPO leading and trailing edges, Laser leading and 
	trailing edges, and Div16 leading and trailing edges.


G0Analysis Release 0.33
     Paul King; 2004/02/19, 19:35

     Mostly minor improvements since the last release.  The highlights are:
     the addition of a Fastbus ntuple branch for the GMS monitor PMTs;
     corection of an array bounds error in the beam scaler pointer routine;
     and correction of a text decoding error in the EPICS event decoding 
     routine.
     Also, a new command line flag was added to force a French configuration
     mode (and ignore the French configuration event); this was a patch to
     allow analysis of runs with porrly formed French configuration events.

     Changes since release 0.32:
     ===========================
     *  Added the cavity "yield" and asymmetry calculation in G0BeamMonitor. 
     *  Updated the DB filling routine WriteToBeamTable(). At this moment, 
	two of the cavity channels are used by the battery signals. So I 
	temporarily tag the measurement_type as 'y' and 'a' for ALL cavity 
	channels. The yield entries should be in the units of V/F bits, and 
	the asymmetry entires are in ppm.
     *  Removed the call of the obsolete routine CavityChargeAsym() in
	Asymmetry.C. 
	Instead incorporate all the asymmetry calculation of the cavity 
	channels in BCMAsym() routine.
     *  Added the NA scaler decoding file for the swapped cables in ROC4.
     *  Added two replayflags.in files, one set up for GH2 analysis,
	and one for LH2 analysis.  They are not selected automatically, but
	need to be explicitly copied over replayflags.in to be used.
     *  Changed (shrink) the range of the halo histogram range to get a better 
	resolution, the range is now 0 to 100 kHz with 1000 bins. 
	at 20 uA, the actual rates are 8kHz with the 11mm hole
     *  Added a branch ("gmsref") containting GMS reference pmt ADCs to
	the fastbus ntuple.
     *  Added a Setup.* file with info on finding the GMS reference PMTs
     *  Correcting a mistaked checkin of G0Error with temporary personal
	edits (I think)
     *  Changed the 2D xy_targ histogram y scale a bit.
     *  Replaced pow() in the VME TDC routines by TMath::Power(). The former 
	complained about some float to double cast and crashed when compiling 
	with gcc3.2.
	Changed the 2D xy_targ histogram y scale a bit.
     *  Some modifications in the graphing routines. Cleaned up the constructor
	to avoid segv.
     *  Found a unit conversion problem when filling the halo rate histogram. 
	The value got filled was actually rates(kHz)/uA*30 ... Fixed the 
	problem. 
	Rebin the halo_rates histogram in g0hists.in from 0 t0 1 kHz/uA for 
	Halo 3.
     *  Added a command line flag --french_cfg to enable specifying French 
	configuration events by hand. This is a fall-back if the French 
	configuration event fails.
	If chosen to fallback on the command line switch to set the French 
	configuration word, do it inside the run loop. Otherwise this word 
	gets reset at beginning of a run!
	Added a test in the command line flag --french_cfg in the main routine 
	such that if we enforce it, it will skip both Get_French_DAC and 
	DecodeFrenchCfg().
     *  Add a charge asymmetry cut (2%) in the quartet cut routine in 
	PassedCuts, to remove the quartet right before the beam trip.
     *  Expanded the scale of the yields of the FPD rings from [0,80] to 
	[0,100] kHz/uA.
     *  Found a bug in this scaler mapping pointer class. The beam data 
	pointer array size was too small (200). Nowadays the beam buffer 
	size has been ~400 words, after the additional VME TDC words.
	I have increase the array size limit to 500.

	Potentially this would cause some nasty memory overloading. I did some 
	comparison with the old and new codes. Looks like all asymmetries 
	computed are still identical so no major harm done.
     *  To avoid confusion in the GUI, now stick to display IHWP In-Out, 
	instead of Out-In. 
	So far, looks like this is the correct sign of the physics asymmetry.
     *  The reason that certain EPICS values in the data stream were not 
	decoded properly is that the "IsNumber" function in StringManip.C 
	was not able to interpret the negative sign.
	I have changed the function to recognize numbers containing "+", 
	"-", "e", and "E", as well as digits and the decimal point.  It 
	is still not very sophisticated, however.


G0Analysis Release 0.34
     Paul King; 2004/02/24, 20:00

     This update principally concerns the realtime analyzer displays,
     but a few things have changed in the main routines as well:
     the clock signal used to normalize the beam scalers is now the
     "YO/4" instead of one of the V/F clock outputs; a database
     filling error was found (the BPM signals had the wrong measurement
     type).

     Changes since release 0.33:
     ===========================
     *  It was discovered that the beam scaler map file decoding is no longer 
	back-ward compatible for runs during Dec. 03, due to the changes 
	the keywords for VME TDC channels. I modified the two older map files
	beam.in.01-23-03, beam.in.11-21-03 -- replacing "BPOYO" by 
	"YO_MUSSON_RISE" and "LASERYO" by "YO_LASER_RISE".
     *  Initialize all ROOT type pointers to zeros in the TIHWPInOut.C c-tor. 
	This is to ensure the d-tor do not generate segv when being called.
     *  Switched to use the YO/4 clock in the G0BeamMonitor class. 
	Redefined the "Yo/4" clock as the "0th" clock in the beam scaler 
	map files (since Jan 04), but after the decoding, scales its 
	reading back as if it were a 4MHz clock. This is to minimize the 
	modifications (no change needed in the feedback code, etc) and to 
	maintain the code backward compatible with the old data, in which 
	we only had one clock signal.
     *  Julie and Kaz pointed out that in the DB beam tables all individual 
	BPM positions have been stored as measurement_type_id = 'y' (should 
	be 'yp').
	Fixed in WriteToSQLDB.C. 
     *  Fixed a bug in scaling the YO/4 clock to mock a 4MHz clock --- 
	scaling factor = 1./(kYO4PerMPS/133332), instead of 
	1./(kYO4PerMPS*133332).
     *  Added a panel in the realtime monitor with a few selected 1D ToF 
	histograms for French, NA and the "hybrid" detectors. Minor cosmetic 
	fix in the realtime monitor code.
     *  Widen the range of the 2D histogram of (x,y) position in g0hists.in.
     *  Decided to display all the "hybrid" ToFs in realtime (only 8).


G0Analysis Release 0.35
     Paul King; 2004/03/02, 14:00

     The main changes are to impose cuts on the Musson box output variables,
     for the 31 MHz and 499 MHz phase shifter control signals, and the
     PLL lock signals to discard events which have incorrectly locked
     beam pick-off timing.  Related to this are the addition of several
     histograms in the real-time display of the Musson box signals and
     the YO timing reference.
     Also added is the new command line flag to set the number of MPS
     to skip after a beam trip; it is currently set to be 2000 by default.

     Changes since release 0.34:
     ===========================
     *  Some cosmetic changes in the Fraser class and the IHWP class ---
	added setters of the x and y axes scales, etc.  Needed for the new 
	goat++ code.
     *  Impose Paul's cut on the YO control voltages in the event-by-event 
	cut to remove the events with bad timing.
     *  Added a command line switch "--tripeventcut <>", to specify the 
	# of MPS to skip after the beam trip recovery. Modify the default 
	replay flag file replayflags.in (skipping 500 MPS by default, 
	according to Kaz's study).
     *  Imposed more cuts on the YO diagonosis signals (31MHz lock, 
	499MHz lock and 1497MHz lock). These cuts are in 
	G0BeamMonitor::YOIsGood() and are applied MPS by MPS.
     *  Added 7 more YO related histograms into realtime displays.
	They are:
	31MHz Ctrl (volts)
	499MHz Ctrl (volts)
	31MHz Lock (volts)
	1497MHz Lock (volts)
	499MHz Lock (volts)
	Aux. YO Select (volts)
	YO_Musson-YO_div16 (ns)
	Modify the histogram binning file g0hists.in.
     *  Fixed a bug in filling the error/alert realtime histogram, which 
	prevented the error counters of detector 16 from displaying. It 
	was due to a mistake in the conversion between fortran-like and 
	c-like array indices.
     *  Put the default MPS cut after trip recovery back to 2000 (70s), 
	just for consistency.


G0Analysis Release 0.36
     Paul King; 2004/03/12, 18:20

     Changed the default beam trip cut to 500 MPS from 2000 MPS.  Added
     "spike" detection routines to the realtime monitor to warn of the
     appearance of the spikes in the French spectra.
     The realtime monitor alarm system has been modified to use both
     audible and visual alarms.
     Several bug fixes were made:  the event cuts for YO errors and
     beam trips; the file segment ordering problem for runs with more
     than 10 segments; the "batch mode" detection for offline analysis.

     Changes since release 0.35:
     ===========================
     *  Changed the default MPS cut after trip recovery to 500 MPS.
	Pursuant to Kaz's recent analysis, the recommendation of the
	analysis/RC's meetings, and the agreement of Ent and Lung.
     *  Fixed a tricky bug in the event by event cut --- now made the 
	YO cut and the beam trip cut in parallel. Otherwise the Yo cut 
	will prevent the 500 MPS counter from incrementing. This 
	should fix the BCM hold off problem observed in the realtime 
	code in version 0.35.
     *  As Louis requested, now also make the same DNL correction on 
	the "buddy" spectra, before filling the French ToFs.
     *  A couple of implementation related to the French spikes. 
	1) Implemented Gilles's spike detection routine into the realtime 
	   monitor. 
	   An audible alarm will be launched if any ToFs show spikes. 
	2) Also made cut on the normalized rate in the FPD rate monitor 
	   (15% tolerance of the variation in the integrated French rate). 
	   Audible alarm implemented.
     *  Fixed the unit of the Yo timing histogram (Musson-Div16).
     *  Need to use the ROOT X11 interface library to make the bell in 
	the background (none xterm mode).  gVirtualX->Bell().
     *  Put a more stringent cut on the event type, so the realtime 
	ToFs and beam 30Hz histograms will only be filled if the event 
	type is 1 and the 120Hz is disabled.
     *  The routine which determines the file segment numbers has been 
	modified to order the file segments in increasing order.  There 
	is no test for missing file segments.
     *  Implemented a better realtime alarm handling --- having the audible 
	beeps as well as an error message box at the bottom of the GUI 
	window. Alarm implemented for French spikes and YO slips. Modified 
	the histogram range a bit.
     *  Modified the "batch mode" test in the offline engine a bit (now 
	test on $DISPLAY and the $JOB_ID variable to decide whether the 
	analyzer is in the batch mode. This should make the analyzer quit 
	if the French configuration event is missing.
     *  Added a small patch to define the Pixel_t data type, for the backward 
	compatibility of the GUI code. Pixel_t type does not seem to be 
	present in the ROOT3.03-06 and earlier versions.


G0Analysis Release 0.37
     Paul King; 2004/03/19, 18:00

     This release only has a few minor changes to the realtime analysis
     to update the Musson box paramter alarms for this running period.
     This also includes changes to decode the two new micropulse counters,
     the number of YO/4 since the last quartet, and the number of YO since
     the last MPS.  These are used in filling the Fastbus ntuples.

     Changes since release 0.36:
     ===========================
     *  Changed the realtime histograms to only fill the YO histograms 
	when the beam current is > 10uA, to remove the potential ambiguity 
	of the control voltage "railing". 
     *  Changed the histogram ranges for the Musson box parameter realtime
	histograms and the "Beam Pickoff"-"Divide by 16" YO timing.
     *  Modified the alarm condition for the time difference between the 
	Beam Pickoff and Divide by 16 YO.  The acceptable range in now 
	+/-0.35 ns from the center of the defined time difference histogram.
     *  Modified the realtime analysis alarm conditions for the 31 MHz control
	and 499 MHz control signals from the Musson box.  The acceptable range 
	is now from 1.0 - 9.0 V for both signals.
     *  Changes to implement the reading of two new hardare scalers:
        1) scaler which counts the number of YO/4 pulses since the last
	   quartet
        2) scaler which counts the number of YO pulses since the last MPS

	Files changed and added:
	1) added beam.in.03-19-2004 with now 6th data word QRT_micro and 
	   13th data word MPS_micro
	2) inputfile_lookup.in to read the new beam.in input file, that 
	   file will be read starting from run 20580 (which is about when 
	   the two signals were added
	3) MapScalerChan.C to implement reading of the new data words which 
	   are stored in fMPSMicro and fQuartetMicro of G0BeamMonitor class
	4) G0BeamMonitor.h G0BeamMonitor.cxx added two variables fMPSMicro 
	   and fQuartetMicro
	5) G0Ntuples.h added new variable in structure FASTBUS_MISC for the 
	   fQuartetMicro value (micro already exist.


G0Analysis Release 0.38
     Paul King; 2004/04/14, 14:55 (Retagged at 17:30)

     The change for which this release was made is to incorporate a new
     Hall A charge asymmetry plot in the realtime histograms.  This
     release also includes:  some improvements to the "Musson box" 
     histograms and alarm; the new command line flag to change the
     "helicity delay" (used for the dummy Pockels cell analysis);
     and a few bug fixes.

     Changes since release 0.37:
     ===========================
     *  Changed the binning of the T_mussion-T_div16 histogram in realtime
	histograms.
     *  Added a "silence" button on the realtime GUI. Minor modifications 
	to the treat the outliners of the YO histograms (if there are
	outliners for more than 300 MPSs, alarm). 
	Changed all YO histograms to log scale.
     *  Added a new HallA charge asymmetry plot on "Parity Quality" in the
	realtime histograms.
     *  Fixed a bug in the superfraser code. Previously it skipped the 
	weighted average on the very last row of the returned values, 
	which caused the  fraser/superfraser discrepancy.
     *  Added a command line option to allow the delay in reporting the 
	helicity to be changed to be changed from being two quartets 
	(eight MPS).  The option is "--helicitydelay #", where # must be 
	a non-negative integer and is the number of quartets which the 
	reported helicity is delayed from the true helicity.
	By changing the value of the helicity delay, it is possible to 
	change the "true" helicity used to caclulate the asymmetries.  
	For instance, a setting of 1 is used to calculate the asymmetries 
	resulting from the influence of the dummy Pockels cell.  A setting 
	of 0 is the same as evaluating the asymmetries using prompt helicity 
	decoding.
     *  Incorrect return function when getting the mean beam parameters in 
	Pass2 of the replay that does the linear regression corrections.  
	This return function prematurely exited the part of the code which 
	reads the mean beam parameter from Pass1.  Bug discovery the courtesy 
	of Silviu.
     *  Minor change to allow use of the new format beam paramter input file
	without the database.  The only change is to the processing of the CLK
	input lines which now have a "channel" and "label".


G0Analysis Release 0.39
     Paul King; 2004/04/26, 17:50

     The change for which this code is being tagged it to include the
     Hall C energy difference in the "parity quality" display of the
     realtime histograms.
     This release also includes:  the new beam cut to eliminate events
     in which iocse20 has frozen; channel decoding files for the Fastbus
     and NA scalers after the NA CFD order was changed; a change in
     histogram (and alarm) limits for the Musson box timing relative
     to the divide-by-16 YO timing; and the addition of a new EPICS
     variable in the event stream.

     Changes since release 0.38:
     ===========================
     *  New cut on beam is included in order to get rid of the IOCSE20
	freezing.
     *  Added the new channel decoding files for the Fastbus and NA 
	scalers to allow for the swapped order of the NA CFD inputs; 
	the first run with the new order is taken as 21516.
     *  Added the location where the raster focuses the beam to the 
	correct size.  EPICS name = EHCFR_LIPRC.  This has also been 
	added to the epics.dictionary.
     *  Changed the histogram limits for the difference between BPO 
	and divide-by-16 YO in realtime analyzer. The nominal difference 
	between the Musson YO and the divide-by-16 YO is now -3.2 ns 
	instead of -3.85 ns.
	The old histogram limits were from -4.625 to -3.075 (with 21 
	bins), and the new histogram limits are -3.975 to -2.425 (with 
	31 bins).
     *  Added HallC energy difference to "parity quality" page of the 
	realtime monitor.


G0Analysis Release 0.40
     Paul King; 2004/05/19, 12:20

     This release includes the final software changes made during the
     forward angle measurement.  These changes were not used by either
     the first-pass or realtime analyzers during the experiment.
     This version will be the baseline for the added features we intend
     to complete over the summer.

     Changes since release 0.39:
     ===========================
     *  Added EPICS variable decoding for each of Halls A, B, and C:
	QE, photocurrent, laser wavelength, and independent transmissions.
	Also added prebuncher amplitude, HallB control and seed levels, 
	laser spot position, and HallA integrated phase noise.
     *  Initialize histogram pointers to NULL in the c-tor.
	Added G0Hists::NullAllHists().  This is to avoid segv when
	calling the d-tor, under the new EL3 gcc/root.


G0Analysis Release 1.01
     Paul King; 2004/07/15, 16:40
     (Retagged at 20:15, 15 July)
     (Retagged at 10:15, 16 July)

     Several changes in the analysis database structure and library
     requirements have been made which break compatibility.
     The "Cut 0" analysis routines have been added.  The PID cuts have
     been changed to include cuts 1, 2, and 3, in addition to the
     pion, proton, inelastic, and background cuts.
     The database access routines have been modified to allow filling
     of data from the Franco-American electronics; although at this
     time the Franco-American data is not calculated or filled in
     the database.
     The blind analysis routines have been added and are enabled for
     all analysis processes.

     The G0Analysis package now requires the OpenSSL library and header
     files are installed with a revision number of at least 0.9.7.

     
     Changes since release 0.40:
     ===========================
     *  Put my private cut0 code into CVS.
	1) Defined a G0Cut0 class, which is derived from G0Scintillator 
	   class. The c-tor initializes a dtimebin[] with size 1 to store 
	   the yield/asymmetry of cut0. All operators defined for 
	   G0Scintillator can be invoked.
	2) Added a G0Cut0 object in G0FrenchDetector. For each MPS, the 
	   counts in cut0 defined on 7 French rings are summed and gets 
	   stored in the G0Cut0 object.
	   Added method G0FrenchDetector::SumCut0().
	3) Calibrated cut0 (each time the timing shifts, cut0 needs to 
	   be recalibrated) and stored the cut files in prminput directory.  
	   Also Modified the lookup file. Modified GetInputData.C and 
	   G0ReplayPrms.C for proper lookup of individual runs.
	4) Went through all operators in G0FrenchDetector class to include 
	   the asymmetry calculation of the cut0 object.
	5) Added cut0 asymmetry and yield into the list of offline histograms.
     *  Merged in Wolfgang's modifications of the PIDs. Instead of defining 
	individual PID data members in G0Scintillator class, defined a 
	vector of struct pidcut[] which contains the cut values, PID name, 
	an integer and double precision numbers to store the yield/asymmetry.
	Modified relevant routines to retrieve the PID cuts from the DB. 
	The code  should now be able to accommodate the additional PID 
	"cut1/2/3".
	It will be a matter of adding a loop to regress on PIDs other than 
	the proton.  Leave it for now.
	Will need to modify the PID calibration scripts to have cut1/2/3 
	calibration properly filled in the DB before the replay.
     *  Added Domi's spike detection routine. If the 1st TST word of the 
	1st DMCH board is not equal to 0xf0f0f0f1, mark all French detectors 
	as bad. A few static constants are defined.
     *  Widened the cut on the G0/G0B BPM wire readouts from 3 bits to 
	5 bits.  This is to ensure that all IOCSE20 crashes are caught.
     *  Minor fix in Domi's spike detection routine -- get the MPS counter 
	right.
     *  The first 100mps is now used to calculate the nominal beam current.  
	The beam trip is seen when the current goes below 90% of this 
	nominal beam current, and the wait-time also starts at 90% of this 
	nominal beam current.
     *  Fixed G0SQL::GetDetectorID() to account for the modified detector
	table structure (octant,detector,make => detector_id), designed to 
	accommodate the hybrid detectors. Changed LoadDetectorTables(), 
	GetScintillatorID() and GetPhotoTubeID().
     *  Added cut0 definition into the DB. Added interface routine to 
	load the cuts.  Currently the code is compute the yield/asymmetry 
	for the following cuts:
	- elastic, pion, inelastics, background, total
	- cut1, cut2, cut3
	- cut0
     *  Modify the DB writing routine to fill the asymmetry/yield of all 
	definedPIDs. Note cut0: (octant = 0, detector = 0, make = fr, 
	detector_id = 1)
     *  Implemented filling/loading of lumi linear regression parameters 
	into/from the DB.
     *  Implemented auto-tagging of the spiky run (data_quality field in 
	the run table). Fixed a bug in the previous version of the French 
	spike detection routine.
     *  Added SumCut0() function call in CalcPIDSum(). This enables the cut0 
	calculation. 
     *  G0 cage temperatures are now in the target_measurements table.
     *  Minor update on the beam cuts.  Beam is tripped if it is below 
	90% of mean current AND less than 4uA.  This ensures that  if a 
	run starts with a beam trip that the ramp-up is not included in 
	the calculation of the mean current.
     *  Added a boolean argument to G0SQL c-tor to choose whether to make 
	a "thin" client or not. If thin==true then do not load 
	detector/monitor/slope_types/monitor_slope_types in the G0SQL 
	constructor. This allows the client like goat++ to switch between 
	gzero_comm and gzero_forward_test (slightly different DB structure).
     *  Added a flag (--seedid <>) to handle specifying the blinding 
	factor seed ID.
     *  Squashed a bug in reading the seed ID from the command line.
     *  Fixed a bug in the lumi regression correction code. Forgot to set 
	the beam parameter means to the right values, so instead of 
	regressing the lumi yields according to the true beam mean values, 
	it regressed them to zeros. 
	Added calls in G0LinearRegression::SetXMean().
     *  Installed the blinded analysis:
	1)  In producing the "blinded" version of G0Analysis, we have 
	    used the OpenSSL library to provide calculation of the MD5 
	    checksums of various strings.
	    The G0Analysis is now dependent on the presence of the
	    OpenSSL include files and library files.  We require at
	    least OpenSSL version 0.9.7.  The OpenSSL libraries appear
	    to be installed in several current Linux distributions;
	    they are installed in /usr/lib and /usr/include on the
	    JLab Enterprise Linux 3 machines.
	    If you must install OpenSSL yourself, or install it in
	    another directory, set the environment variable
	    OPENSSL_DIR to the directory in which you have it
	    installed.  The makefile should recognize this variable
	    and do the compilation properly.
	    If you are unable to use OpenSSL, and would like to
	    suggest another library for computation of MD5 sums,
	    please let me know (at this time, the ROOT TMD5 class
	    seems to not match "standard" MD5 calculations, and the
	    algorithm may change in future ROOT releases; that's why
	    we're not using it).
	2)  The routine "CalcAsyms" has been encapsulated in the new 
	    class G0Blinder.
	    This new class contains the code to compute and apply the 
	    blinding factor to the detector asymmetries and the detector 
	    helicity correlated differences.
	    The seed string will normally be extracted from the G0Database.  
	    At the conclusion of the analysis, the MD5 checksum of the 
	    blinding factor and ten "test" blinded variables will be 
	    written into the database.
	    The Helicity and Asymmetry classes now have a "Scale(scale)" 
	    member function which uses "Normalize(1.0/scale)" member 
	    functions of G0NADetector and G0FrenchDetector classes.
     *  Modified the G0Realtime source and include files to support the 
	use of the new blinding analysis routines.
     *  Minor modifications in the G0Blinder class to standardize the 
	calculation of the blinding factor and the test values.  Also fixed 
	a few bugs in the database filling routines.
     *  =====
	Retagged at 20:15, 15 July, because:
     	    Another set of bugs were discovered in the blinding rouitnes just
	    after tagging the code.  These were due to the fact that 
	    "Normalize" is set up to not effect the Cut0 object within the 
	    G0FrenchDetector class.
     *  =====
	Retagged at 10:15, 16 July, because:
	    Commented out two lines for debugging of the database filling
	    routines for the test values.


G0Analysis Release 1.02
     Paul King; 2004/12/15, 15:20

       There have been some minor changes in the G0Blinder routine:  a constant
     was added to one of the private functions to correct a compilation
     error; a public function "PrintFinalValues()" was added to display
     the blinding factor checksum and blinded test values for analysis jobs
     that are not outputting to the database; the "test parameters" written
     into the database were not blinded, that has been corrected, and the
     "test vale" seed has been modified so that we cannot determine the
     blinding factor by comparing the old and new test values.
       The G0Buddies class was modifed to calculate the Figure of Merit, error
     and sigma of the buddy counters.  These buddy values are not yet stored
     in the database.
       Some changes and bug fixes in  the linear regression code:  when trying
     to apply the linear regression correction, if a NULL value is returned
     when filling the linear regression parameters from the database, then the
     internal variable will be set to zero, if all values are NULL, the run will
     exit with a warning message; bug correction to resolve a regression to zero;
     added the pion, background, and total cuts to the linear regression scheme.
       Changed the the function calls and variable definitions for "G0BeamMonitor"
     variables in the several classes into "G0TriggerSupervisor" variables to
     support the upcoming changes for the YO analysis.
       Added a test on the database account and access mode before trying to 
     write the new calib_id into the database for runs which require a new 
     set of calibration parameters; if the account does not have write permission
     to the database, a message will be printed explaining what entries need to
     be added to the database by hand.


     Changes since release 1.01:
     ===========================
     *  Modified a 64-bit constant definition in one of the G0Blinder private
	functions to avoid a compilation error with GCC version 3.3 and above.
	Also added a public member function "PrintFinalValues()" to allow
	display of the blinding factor checksum and blinded test values
	for analysis jobs that are not outputting to the database.
     *  Added FoM, Error and sigma variables and the related calculations to 
	the G0Buddies class in order to implement later the routine to fill 
	the database. Before these numbers were calculated and filled directly 
	into histograms in the G0Hists class. The various calculations were 
	removed from the G0Hists class, where now the histograms are filled 
	using the numbers of the G0Buddies class.
     *  Added a test on NULL values returned by databse queries when loading the
	linear regression parameters from the database.  If a database entry
	is NULL, the internal variable for the detector slope, lumi slope, or
	mean beam parameter value will be set to zero.
	If there is not at least one valid detector slope, lumi slope, and mean
	beam parameter value for the run, the analysis will exit with a warning
	message.
	This only applies for analyses in which we are applying the linear
	regression corrections.
     *  Bug report.  The return statement on the fetching of the beam parameters 
	were STILL in the wrong place.  Effectively, we were regressing to zero 
	for some of the beam paramters.
     *  Corrected an error in the blinding routine; the "test values" output to the
	database at the end of each run were not blinded.  The output "test values"
	are now the blinded ones.  The seed string for the test values has been
	changed, so the test values will be different than those produced with
	the first version of the blinding code.
     *  Modified the G0SQL::SelectCalibrationID function to add a test on the
	database account and access mode before trying to write the new calib_id
	into the database for runs which require a new set of calibration parameters.
	If the database account does not have write permission to the database
	a message will be printed explaining what entries need to be added to
	the database by hand.
     *  I have changed the function calls and variable definitions for "G0BeamMonitor"
	variables in the following classes into "G0TriggerSupervisor" variables.
	This chage is to support the YO analysis changes which will be submitted
	later.
		AssignHelicity, Asymmetry, G0Blinder, G0LinearRegression,
		Helicity, WriteToSQLDB
     *  Added pion, background, and total cuts to the linear regression scheme.


G0Analysis Release 1.03
     Paul King; 2004/12/22, 14:30

     This release includes the merged code from branch release G0Analysis_0_31_01_e.
     This noe contains the FrancoAmerican analysis routines.  This version also
     has the updated linear regression routines to allow regression corrections
     on each of the particle ID cuts.

     Changes since release 1.02:
     ===========================
     *  Added cut1, cut2, and cut3 to the linear regression scheme.  The linear
	regression scheme is now fairly extensible if we want to add more cuts
	later in the ananlysis.
     *  Made a series of corrections to the fFramFPD indices in 
	G0LinearRegression.h and G0LinearRegression.C, and re-enabled the 
	FrAm slope output routine in WriteToSQLDB.C.  The FrAm linear regression 
	slopes now appear to be properly output to the database; the problem 
	had been caused by an attempted out-of-range array access.
     *  Updates from the G0Analysis branch version 0_31_01_e:
	*  Added a fFrAm flag in G0CodaEvent, VCodaEvent. Test will be 
	   made on this flag before further processing the calculation.
	*  Added a fJumpFront2Back variable in G0FrenchDetector and 
	   G0FrAmDetector. 
	   This variable will handle the indexing of Front and Back 
	   scintillator hits.  Made relevant changes in the computation 
	   routines  G0FrenchDetector::+,-,+=,-=,=,MeanValue(),
	   SigmaValue(), etc, and the SDMCH related routines. Modify 
	   the SDMCH array assignment (before there was an indexing 
	   error related to the back hits). The FrAm detector asymmetry
	   computation should be using the same methods in G0FrenchDetector 
	   class.
	*  Modified Helicity and Asymmetry class to contain the FrAm 
	   detector class.  Modified AssignHelicity to fill the FrAm 
	   values into the quartet buckets.  Also added a flag in 
	   quartet-analysis structure, for potential switching-off 
	   of the FrAm analysis branch. Modified G0Cuts class and the 
	   CutOnDetectors routine, to incorporate the 10 sigma yield 
	   cuts calculation of the FrAm detector. 
	*  Some relevant modifications in the main routines G0Analysis.C 
	   and G0Realtime.C.
        *  The transition between G0Analysis_0_31_01_a and 
	   G0Analysis_0_31_01_b had too many detailed changes to 
	   describe here.
	   See the change log in branch G0Analysis_0_31_01 or use
	   "cvs diff -r G0Analysis_0_31_01_a -r G0Analysis_0_31_01_b"
	   to see the changes.
        *  The transition between G0Analysis_0_31_01_b and 
	   G0Analysis_0_31_01_c had too many detailed changes to 
	   describe here.
	   See the change log in branch G0Analysis_0_31_01 or use
	   "cvs diff -r G0Analysis_0_31_01_b -r G0Analysis_0_31_01_c"
	   to see the changes.
        *  The transition between G0Analysis_0_31_01_c and 
	   G0Analysis_0_31_01_d had too many detailed changes to 
	   describe here.
	   See the change log in branch G0Analysis_0_31_01 or use
	   "cvs diff -r G0Analysis_0_31_01_c -r G0Analysis_0_31_01_d"
	   to see the changes.
	*  Modified G0FrAmDetectors::LoadDNL() and LoadPID() to load 
	   the calibrations from DB correctly. Minor modifications to NA 
	   and French class to remove the warning messages when loading 
	   the PID cuts.
	*  Discovered the size of the obj array *fFramFPD[][] was declared 
	   incorrectly.  The regression implementation assumed [4][16], 
	   but it was defined as [4][2].  Fixing this resolved the segv 
	   when deleting g0cuts objects.
	*  Found the possible cause of the memory overriding problem 
	   ("corrupted string in g0sql object" found by Sarah and Paul). 
	   The size of the vector to store the FrAm values are not adequately 
	   allocated: all code implementation assumed [4][16] and the vectors 
	   were initialized to [4][2].  This happens to both the mps and 
	   quartet ntuple. Fixing it seemed to get rid of the problem, 
	   although more tests are needed to verify this.
	*  Several minor changes to G0FrenchDetector.C, G0Globals.C, 
	   G0Ntuples.C, G0SQL.C, and G0Analysis.C  to remove debugging 
	   comments and otherwise remove any non-important differences 
	   between this branch and the main trunk, version G0Analysis_1_02.


G0Analysis Release 1.04
     Paul King; 2005/01/07, 17:50

     This release contains the new YO TDC analysis routines to calculate
     and store in the G0 database the run-averages helicity averaged time 
     and helicity correlated time differences for the different YO sources.
     It also now will store the NA buddy Figure of Merit and errors.

     Changes since release 1.03:
     ===========================
     *  These are the changes in the YO TDC analysis to caclulate and 
	store in the database the run-averages helicity averaged time 
	and helicity correlated time differences.  New "monitors" and 
	"measurement_types" entries have been made to label these data; 
	the data are stored in the "beam" table.
	These routines are compatable with the gzero_forward_test2 
	database, and will be compatable with the next pass database; 
	some errors may occur if using an older database with this 
	analyzer, but it should continue to work for database retrieval 
	operations.
     *  Fixed a bug which prevented the 2nd pass from actually applying 
	the linear regression corrections.  I think this was a pretty 
	recent bug, since the corrections were working properly at the 
	beginning of 12/2004.  The problem was that a conditional 
	statement which checked whether the make of a particular 
	detector was FR or NA was not functioning properly in the 
	routine SetSlopes().  This caused all the slopes loaded from 
	the DB at the beginning of the 2nd pass to be zero (hence no 
	correction was applied).  
     *  Minor changes in G0Analysis and G0CodaEvent so that the FrAm 
	configuration event decoding routines are run only if there 
	has been a FrAm configuration buffer in the French configuration 
	event.  This is required for compatability with runs without the 
	FrAm DMCH installed.
     *  Modifed the beam.in files for the second commisioning run and the
	production run to match the new format for the VME TDC labelling.
	Also added a new file "beam.in.01-20-2004_b" which I think is valid
	from about run 19171 until 19238; the VME TDC numbers in the database
	should be checked for jumps once we begin refilling the database.
	Large shifts might indicate an incorrect mapping for particular
	runs.
     *  Added a function to the G0Buddies class to write the Figure Of Merit,
	sigma, and error for the NA buddy counters into the database.  There
	has been a little confusion about the definitions of "Sig" and "Err"
	between the analyzer and the database, but I have decided now not
	to change the definitions in the analyzer; the definitions are:
          Sig     The calculated error on the mean of the Figure of Merit.
          Err     The fitted mean of the FoM errors.
	Both of these are scaled by sqrt(NumQRTs).
	The database documentation should be updated to reflect these 
	definitions at some point.
     *  Minor correction to the G0Buddies MySQL access routine; the 
	function needs to explicitly open and close the connection to 
	the database.