rivet 1.8.3-1.1 / usr / bin / rivet-rescale

#!/usr/bin/env python

"""\
%prog [-r <REFDATAPATH>] [-O <observable-file>] [-b <bindef> [-b ...]] <AIDAFILE> [<OUTFILE>]

Rescale histos in observable-file of AIDAFILE to the area of the
corresponding histos in REFDATAPATH. REFDATAPATH can either be
a single AIDA-file or a directory containing AIDA-files. By default
the standard Rivet reference files are used.

Observable definitions of the form
/CDF_2000_S4155203/d01-x01-y01  1.0
can be used to specify an absolute normalisation (1.0, here) for a
histogram rather than using the reference histo. If the --multiply
switch is used, the ref histo area will be scaled by the given factor
to give the target normalisation.

You can also define bins to chop out in the observable-file in the
same way as in chop_bins:
/CDF_2000_S4155203/d01-x01-y01:0:35  1.0
This will chop the bins with Z-pT > 35 GeV and obtain a rescaling factor
from that restricted bin range: note that the output histograms will
be rescaled but *unchopped*.

Only one bin definition can be used for each histogram, so the last bindef
specified for that histo path will be the one which is applied. The bindefs
are constructed in order from those in the obsfile, and then those given on
the command line with the -b flag, like this:
    -b "/CDF_2000_S4155203/d01-x01-y01:5:135 2.0"


Examples:

 * %prog out.aida
   This will return the histos in out.aida, scaled to match the overall
   normalisations of the Rivet ref data.

 * %prog -O observables.obs out.aida
   This will return the histos in out.aida, scaled by the bin definitions
   specified in observables.obs (and using the Rivet ref data again)

 * %prog -r path/to/CDF_2000_S4155203.aida \
     -b "/CDF_2000_S4155203/d01-x01-y01:2:5" out.aida
   For this Z-boson pT-distribution, the normalisation to the provided ref
   data file is only applied between 2 < x < 5 GeV.
"""

import sys
if sys.version_info[:3] < (2,4,0):
    print "rivet scripts require Python version >= 2.4.0... exiting"
    sys.exit(1)

import os, re, logging
from lighthisto import Histo

# try:
#     from IPython.Shell import IPShellEmbed
#     ipshell = IPShellEmbed([])
# except:
#     logging.info("Ipython shell not available.")


## Try to load faster but non-standard cElementTree module
try:
    import xml.etree.cElementTree as ET
except ImportError:
    try:
        import cElementTree as ET
    except ImportError:
        try:
            import xml.etree.ElementTree as ET
        except:
            sys.stderr.write("Can't load the ElementTree XML parser: please install it!\n")
            sys.exit(1)


def getHistosFromAIDA(aidafile):
    '''Get a dictionary of histograms indexed by name.'''
    if not re.match(r'.*\.aida$', aidafile):
        logging.debug("Error: input file '%s' is not an AIDA file" % aidafile)
    aidafilepath = os.path.abspath(aidafile)
    if not os.access(aidafilepath, os.R_OK):
        logging.debug("Error: cannot read from %s" % aidafile)

    histos = {}
    tree = ET.parse(aidafilepath)
    for dps in tree.findall("dataPointSet"):
        ## Get this histogram's path name
        dpsname = os.path.join(dps.get("path"), dps.get("name"))
        ## Is it a data histo?
        h = Histo.fromDPS(dps)
        h.isdata = dpsname.upper().startswith("/REF")
        if h.isdata:
            dpsname = dpsname.replace("/REF", "")
        histos[dpsname] = h
    return histos


def getRefHistos(refpaths, analyses):
    """
    Return dictionary of reference histos {name: histo}.
    refpaths can either be a single file or a directory.
    """
    refhistos = {}
    if refpaths:
        for refpath in refpaths:
            if os.path.isfile(refpath):
                logging.debug("Reading ref histos from file %s" % refpath)
                refhistos = getHistosFromAIDA(refpath)
            elif os.path.isdir(refpath):
                if opts.fast:
                    logging.debug("Fast mode - not loading all data-files")
                    for ana in analyses:
                        refaida = os.path.join(refpath, ana+".aida")
                        if os.path.isfile(refaida):
                            temp = getHistosFromAIDA(refaida)
                            for k, v in temp.iteritems():
                                if not k in refhistos.keys():
                                    refhistos[k] = v
                else:
                    for aida in os.listdir(refpath):
                        if aida.endswith(".aida"):
                            temp = getHistosFromAIDA(os.path.join(refpath, aida))
                            for k, v in temp.iteritems():
                                if not k in refhistos.keys():
                                    refhistos[k] = v
                logging.debug("Read ref histos from folder %s" % refpath)
    return refhistos


def readObservableFile(obsfile):
    """ Read observables to normalise from file obsfile.
        Return-values are a list of the histo-names to normalise and a
        dictionary with name:newarea entries.
    """
    obslist = set()
    obsnorms = {}
    bindefs = {}

    if obsfile is not None:
        try:
            f = open(obsfile, 'r')
        except:
            logging.error("Cannot open histo list file %s" % opts.OBSFILE)
            sys.exit(2)
        for line in f:
            stripped = line.strip()
            # Skip empty or commented lines
            if len(stripped) == 0 or stripped.startswith("#"):
                continue

            # Split the line to find out whether newarea is given in obsfile
            path, low, high, area = getBindef(line)
            bindefs[path] = (low, high)
            if area:
                obsnorms[path] = float(area)

            obslist.add(path)
        f.close()
    return obslist, obsnorms, bindefs


def getBindef(line):
    """ Try to read bin definitions (xlow, xhigh, newarea) from single
        string.
    """
    area = None
    # Try to read bin specs for chopping
    # Remove possible comments at the end of the line
    splitline = line.strip().rsplit("#")[0].split()
    try:
        path, low, high = splitline[0].split(":")
    except:
        path = splitline[0].split(":")[0]
        low  = ""
        high = ""
        logging.debug("No bin range supplied for %s" % path)
    low = float(low) if low else None
    high = float(high) if high else None

    # Try to get area to normalise to
    logging.debug("Trying to read area to rescale to from bindef...")
    if len(splitline) == 2:
        try:
            area = float(splitline[1])
            logging.debug("Success: %e" % area)
        except:
            logging.debug("Failed: %s" % splitline[1])
            pass
    return (path, low, high, area)



if __name__ == "__main__":
    from optparse import OptionParser, OptionGroup
    parser = OptionParser(usage=__doc__)

    parser.add_option("-O", "--obsfile", dest="OBSFILE", default=None,
                      help="Specify a file with histograms (and bin ranges) that are to be normalised.")
    parser.add_option("-b", "--bins", dest="BINRANGES", action="append", default=[],
                      help="Specify a histogram and bin range that is to be used. The format is `AIDAPATH:start:stop'.")
    parser.add_option("-r", "--refdir", dest="REFDIR", default=None,
                      help="File of folder with reference histos")
    parser.add_option("-a", dest="AIDA", default=True, action="store_true",
                      help="Produce AIDA output rather than FLAT")
    parser.add_option("-f", dest="AIDA", default=True, action="store_false",
                      help="Produce FLAT output rather than AIDA")
    parser.add_option("--multiply", dest="MULTIPLY", default=False, action="store_true",
                      help="Rescale histos using weight given as factor rather than new area")
    parser.add_option("-i", "--in-place", dest="IN_PLACE", default=False, action="store_true",
                      help="Overwrite input file rather than making input-rescaled.aida")
    parser.add_option("--fast", default=False, action="store_true",
                      help="Try loading only reference files from refpath that match analyses in input-file" )
    verbgroup = OptionGroup(parser, "Verbosity control")
    verbgroup.add_option("-v", "--verbose", action="store_const", const=logging.DEBUG, dest="LOGLEVEL",
                         default=logging.INFO, help="print debug (very verbose) messages")
    verbgroup.add_option("-q", "--quiet", action="store_const", const=logging.WARNING, dest="LOGLEVEL",
                         default=logging.INFO, help="be very quiet")
    opts, args = parser.parse_args()

    ## Configure logging
    logging.basicConfig(level=opts.LOGLEVEL, format="%(message)s")

    ## Check number of args
    if len(args) not in (1, 2):
        print "Usage: %s" % __doc__.splitlines()[0]
        sys.exit(1)

    ## Get MC histos
    histos = getHistosFromAIDA(args[0])

    # Get unique analyses identifiers to speed up ref-data loading
    analyses = set([obs.split("/")[1] for obs in histos.keys()])

    # Read in reference histos to get reference areas to normalise to
    refdirs = []
    if opts.REFDIR:
        refdirs.append(opts.REFDIR)
    else:
        import rivet
        refdirs += rivet.getAnalysisRefPaths()
    refhistos = getRefHistos(refdirs, analyses)
    if len(refhistos) == 0 and not opts.MULTIPLY:
        logging.warning("You haven't specified any reference histograms. You'd better know what you're doing!")

    # Read in observables, if no bindefinitions are given in the file or the
    # command line, all observables will be renormalised if possible to the
    # corresponding refhisto area
    obslist, obsnorms, bindefs = readObservableFile(opts.OBSFILE)
    if opts.BINRANGES:
        for b in opts.BINRANGES:
            name = b.strip().split(":")[0]
            path, low, high, area = getBindef(b)
            obslist.add(name)
            bindefs[name] = (low, high)
            if area:
                obsnorms[name] = area
    if len(obslist) == 0 and not opts.BINRANGES:
        logging.info("No bin-definitions given: all histos will be rescaled to match the data")
        obslist = histos.keys()

    ## Create output filename
    base = args[0].split(".aida")[0]
    if len(args) > 1:
        outfile = args[1]
    else:
        if not opts.IN_PLACE:
            base += "-rescaled"
        if opts.AIDA:
            outfile = base + ".aida"
        else:
            outfile = base + ".dat"

    aidaheader = """<?xml version="1.0" encoding="ISO-8859-1" ?>
<!DOCTYPE aida SYSTEM "http://aida.freehep.org/schemas/3.3/aida.dtd">
<aida version="3.3">
  <implementation version="1.1" package="Rivet"/>
    """
    # Open file for output
    f = open(outfile, "w")
    if opts.AIDA:
        f.write(aidaheader)


    # Iterate over all histos
    for name, histo in histos.iteritems():
        # Don't normalise all histos found
        if name in obslist:
            # Find out whether ref-histo is given
            if name in refhistos.keys():
                logging.debug("Rescaling to ref-histo for %s" % name)
                tempref = refhistos[name]
            else:
                logging.debug("Not using refhisto for rescaling of %s" % name)
                tempref = histos[name]

            # Try to chop bins
            if name in bindefs.keys():
                logging.debug("Using bindefs for rescaling of %s" % name)
                tempref = tempref.chop(bindefs[name])
                tempold = histos[name].chop(bindefs[name])
            else:
                logging.debug("Not using bindefs for rescaling of %s" % name)
                #tempref = refhistos[name]
                tempold = histos[name]

            # Get old and new histogram areas
            oldarea = tempold.getArea()
            if name in obsnorms.keys():
                # Check if we want to scale histos by a factor
                if opts.MULTIPLY:
                    newarea = oldarea*obsnorms[name]
                else:
                    # Rescale to manually given new area
                    newarea = obsnorms[name]
            # Rescale this histo to ref-histo area
            else:
                newarea = tempref.getArea()

            if not oldarea == 0.0:
                scalefactor = newarea/oldarea
            else:
                scalefactor= 0.0
            if scalefactor != 1.0 and scalefactor > 0.0:
                oldarea = histos[name].getArea()
                newarea = histos[name].getArea() * scalefactor

                # Scale histo
                logging.info("Rescaling %s by factor %.3e (area %.3e -> %.3e)" % (name, scalefactor, oldarea, newarea))
                renormed = histos[name].renormalise(newarea)

                # Write to file
                if opts.AIDA:
                    f.write(renormed.asAIDA())
                else:
                    f.write(renormed.asFlat())
                continue

        ## Fallback to no rescaling if loop is not escaped early
        if opts.AIDA:
            f.write(histos[name].asAIDA())
        else:
            f.write(histos[name].asFlat())


    if opts.AIDA:
        f.write("</aida>")

    logging.debug("Output written to %s" % outfile)