+ return total,outfilename
+
+
+def do_midx(outdir, outfilename, infilenames):
+ rv = _do_midx(outdir, outfilename, infilenames)
+ if rv:
+ print rv[1]
+
+
+def do_midx_dir(path):
+ already = {}
+ sizes = {}
+ if opt.force and not opt.auto:
+ midxs = [] # don't use existing midx files
+ else:
+ midxs = glob.glob('%s/*.midx' % path)
+ contents = {}
+ for mname in midxs:
+ m = git.open_idx(mname)
+ contents[mname] = [('%s/%s' % (path,i)) for i in m.idxnames]
+ sizes[mname] = len(m)
+
+ # sort the biggest midxes first, so that we can eliminate smaller
+ # redundant ones that come later in the list
+ midxs.sort(lambda x,y: -cmp(sizes[x], sizes[y]))
+
+ for mname in midxs:
+ any = 0
+ for iname in contents[mname]:
+ if not already.get(iname):
+ already[iname] = 1
+ any = 1
+ if not any:
+ log('%r is redundant\n' % mname)
+ unlink(mname)
+ already[mname] = 1
+
+ midxs = [k for k in midxs if not already.get(k)]
+ idxs = [k for k in glob.glob('%s/*.idx' % path) if not already.get(k)]
+
+ for iname in idxs:
+ i = git.open_idx(iname)
+ sizes[iname] = len(i)
+
+ all = [(sizes[n],n) for n in (midxs + idxs)]
+
+ # FIXME: what are the optimal values? Does this make sense?
+ DESIRED_HWM = opt.force and 1 or 5
+ DESIRED_LWM = opt.force and 1 or 2
+ existed = dict((name,1) for sz,name in all)
+ log('%d indexes; want no more than %d.\n' % (len(all), DESIRED_HWM))
+ if len(all) <= DESIRED_HWM:
+ log('Nothing to do.\n')
+ while len(all) > DESIRED_HWM:
+ all.sort()
+ part1 = [name for sz,name in all[:len(all)-DESIRED_LWM+1]]
+ part2 = all[len(all)-DESIRED_LWM+1:]
+ all = list(do_midx_group(path, part1)) + part2
+ if len(all) > DESIRED_HWM:
+ log('\nStill too many indexes (%d > %d). Merging again.\n'
+ % (len(all), DESIRED_HWM))
+
+ for sz,name in all:
+ if not existed.get(name):
+ print name
+
+
+def do_midx_group(outdir, infiles):
+ for sublist in _group(infiles, opt.max_files):
+ rv = _do_midx(path, None, sublist)
+ if rv:
+ yield rv
+
+