def main():
args = utils.get_args('generate pub-years data')
outf = utils.open_output()
weekdays = [ 'Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun' ]
pubyears = {} # set()
for xd in xdfile.corpus():
puby = (xd.publication_id(), xd.year())
if puby not in pubyears:
pubyears[puby] = []
pubyears[puby].append(xd)
if pubyears:
metadb.delete_stats()
for puby, xdlist in sorted(pubyears.items()):
pubid, year = puby
npublic = 0
# TODO: SELECT FROM publications
nexisting = 0
# organize by day-of-week
byweekday = {}
byweekday_similar = {}
for w in weekdays:
byweekday[w] = []
byweekday_similar[w] = []
for xd in xdlist:
dow = dow_from_date(xd.get_header('Date'))
if dow: # Might be empty date or only a year
byweekday[dow].append(xd)
for r in metadb.xd_similar(pubid + str(year)):
if r.match_pct < 25:
continue
xd = xdfile.get_xd(r.xdid)
if xd:
dt = xd.get_header('Date')
if dt:
dow = dow_from_date(dt)
if dow: # Might be empty date or only a year
byweekday_similar[dow].append(r)
else:
debug("Date not set for: %s" % xd)
# tally stats
for weekday in weekdays:
copyrights = Counter() # [copyright_text] -> number of xd
editors = Counter() # [editor_name] -> number of xd
formats = Counter() # ["15x15 RS"] -> number of xd
# todo
nexisting = 0
nxd = len(byweekday[weekday])
public_xdids = [] # Empty for now
for xd in byweekday[weekday]:
xdid = xd.xdid()
if (year.isdigit() and int(year) <= 1965) or xdid in public_xdids:
npublic += 1
editor = xd.get_header('Editor').strip()
if editor:
editors[editor] += 1
sizestr = xd.sizestr()
if sizestr:
formats[sizestr] += 1
copyright = xd.get_header('Copyright').strip()
if copyright:
copyrights[copyright] += 1
# debug("ME: %s MCPR: %s MF: %s" % (list(editors), list(copyrights), list(formats)))
def process_counter(count, comp_value):
# Process counter comparing with comp_value
if count:
item, num = count.most_common(1)[0]
if num != comp_value:
item += " (%s)" % num
else:
item = ''
return item
#
maineditor = process_counter(editors, nxd)
maincopyright = process_counter(copyrights, nxd)
mainformat = process_counter(formats, nxd)
reprints = 0
touchups = 0
redones = 0
copies = 0
themecopies = 0
for r in byweekday_similar[weekday]:
xd1 = xdfile.get_xd(r.xdid)
xd2 = xdfile.get_xd(r.match_xdid)
if xd1 is None:
info("%s: similar puzzle %s not in corpus" % (r.match_xdid, r.xdid))
continue
if xd2 is None:
info("%s: similar puzzle %s not in corpus" % (r.xdid, r.match_xdid))
continue
dt1 = xd1.get_header('Date')
dt2 = xd2.get_header('Date')
aut1 = xd1.get_header('Author').lower()
aut2 = xd2.get_header('Author').lower()
pct = int(r.match_pct)
if dt2 < dt1: # only capture the later one
##deduce_similarity_type
if diff_authors(aut1, aut2): # suspicious
if pct >= 50:
copies += 1
elif pct >= 30:
themecopies += 1
else:
if pct == 100:
reprints += 1
elif pct >= 50:
touchups += 1
elif pct >= 30:
themecopies += 1
metadb.append_row("pub/stats",
(pubid, year, weekday,
mainformat, maineditor, maincopyright,
nexisting, nxd, npublic,
reprints, touchups, redones,
copies, themecopies))
def main():
global args
parsers = {
'.xml': [parse_ccxml, parse_uxml],
'.json': [parse_ujson],
'.puz': [parse_puz],
'.html': [parse_xwordinfo],
'.pdf': [],
'.jpg': [],
'.gif': [],
'.xd': [], # special case, just copy the input, in case re-emitting screws it up
}
p = args_parser('convert crosswords to .xd format')
p.add_argument('--copyright', default=None, help='Default value for unspecified Copyright headers')
p.add_argument('--extsrc', default=None, help='Value for receipts.ExternalSource')
p.add_argument('--intsrc', default=None, help='Value for receipts.InternalSource')
p.add_argument('--pubid', default=None, help='PublicationAbbr (pubid) to use')
args = get_args(parser=p)
outf = open_output()
for input_source in args.inputs:
try:
# collect 'sources' metadata
source_files = {}
# collect receipts
receipts = []
for fn, contents, dt in find_files_with_time(input_source, ext='.tsv'):
progress(fn)
for row in parse_tsv_data(contents.decode('utf-8'), "Source"):
innerfn = strip_toplevel(row.SourceFilename)
if innerfn in source_files:
warn("%s: already in source_files!" % innerfn)
continue
source_files[innerfn] = row
# enumerate all files in this source, reverse-sorted by time
# (so most recent edition gets main slot in case of shelving
# conflict)
for fn, contents, dt in sorted(find_files_with_time(input_source, strip_toplevel=False), reverse=True, key=lambda x: x[2]):
if fn.endswith(".tsv") or fn.endswith(".log"):
continue
if not contents: # 0-length files
continue
innerfn = strip_toplevel(fn)
if innerfn in source_files:
srcrow = source_files[innerfn]
CaptureTime = srcrow.DownloadTime
ExternalSource = args.extsrc or srcrow.ExternalSource
SourceFilename = innerfn
else:
debug("%s not in sources.tsv" % innerfn)
CaptureTime = iso8601(dt)
ExternalSource = args.extsrc or parse_pathname(input_source).filename
SourceFilename = innerfn
ReceivedTime = iso8601(time.time())
InternalSource = args.intsrc or parse_pathname(input_source).filename
already_received = metadb.check_already_received(ExternalSource, SourceFilename)
xdid = ""
prev_xdid = "" # unshelved by default
existing_xdids = set(r.xdid for r in already_received)
if existing_xdids:
if len(existing_xdids) > 1:
warn('previously received this same file under multiple xdids:' + ' '.join(existing_xdids))
else:
prev_xdid = existing_xdids.pop()
debug('already received as %s' % prev_xdid)
# try each parser by extension
ext = parse_pathname(fn).ext.lower()
possible_parsers = parsers.get(ext, parsers[".puz"])
progress(fn)
if ext == ".xd":
outf.write_file(fn, contents.decode('utf-8'), dt)
elif not possible_parsers:
rejected = "no parser"
else:
rejected = ""
for parsefunc in possible_parsers:
try:
try:
xd = parsefunc(contents, fn)
except IncompletePuzzleParse as e:
error("%s %s" % (fn, e))
xd = e.xd
if not xd:
continue
xd.filename = replace_ext(strip_toplevel(fn), ".xd")
if not xd.get_header("Copyright"):
if args.copyright:
xd.set_header("Copyright", args.copyright)
catalog.deduce_set_seqnum(xd)
xdstr = xd.to_unicode()
mdtext = "|".join((ExternalSource,InternalSource,SourceFilename))
xdid = prev_xdid or catalog.deduce_xdid(xd, mdtext)
path = catalog.get_shelf_path(xd, args.pubid, mdtext)
outf.write_file(path + ".xd", xdstr, dt)
rejected = ""
break # stop after first successful parsing
except xdfile.NoShelfError as e:
error("could not shelve: %s" % str(e))
rejected += "[shelver] %s " % str(e)
except Exception as e:
error("%s could not convert [%s]: %s" % (parsefunc.__name__, fn, str(e)))
rejected += "[%s] %s " % (parsefunc.__name__, str(e))
# raise
if rejected:
error("could not convert: %s" % rejected)
# only add receipt if first time converting this source
if already_received:
debug("already received %s:%s" % (ExternalSource, SourceFilename))
else:
receipts.append([
CaptureTime,
ReceivedTime,
ExternalSource,
InternalSource,
SourceFilename,
xdid
])
for r in receipts:
metadb.append_row('gxd/receipts', r)
except Exception as e:
error(str(e))
if args.debug:
raise
def main():
global args
parsers = {
'.xml': [parse_ccxml, parse_uxml],
'.json': [parse_ujson],
'.puz': [parse_puz],
'.html': [parse_xwordinfo],
'.pdf': [],
'.jpg': [],
'.gif': [],
'.xd':
[], # special case, just copy the input, in case re-emitting screws it up
}
p = args_parser('convert crosswords to .xd format')
p.add_argument('--copyright',
default=None,
help='Default value for unspecified Copyright headers')
p.add_argument('--extsrc',
default=None,
help='Value for receipts.ExternalSource')
p.add_argument('--intsrc',
default=None,
help='Value for receipts.InternalSource')
p.add_argument('--pubid',
default=None,
help='PublicationAbbr (pubid) to use')
args = get_args(parser=p)
outf = open_output()
for input_source in args.inputs:
try:
# collect 'sources' metadata
source_files = {}
# collect receipts
receipts = []
for fn, contents, dt in find_files_with_time(input_source,
ext='.tsv'):
progress(fn)
for row in parse_tsv_data(contents.decode('utf-8'), "Source"):
innerfn = strip_toplevel(row.SourceFilename)
if innerfn in source_files:
warn("%s: already in source_files!" % innerfn)
continue
source_files[innerfn] = row
# enumerate all files in this source, reverse-sorted by time
# (so most recent edition gets main slot in case of shelving
# conflict)
for fn, contents, dt in sorted(find_files_with_time(
input_source, strip_toplevel=False),
reverse=True,
key=lambda x: x[2]):
if fn.endswith(".tsv") or fn.endswith(".log"):
continue
if not contents: # 0-length files
continue
innerfn = strip_toplevel(fn)
if innerfn in source_files:
srcrow = source_files[innerfn]
CaptureTime = srcrow.DownloadTime
ExternalSource = args.extsrc or srcrow.ExternalSource
SourceFilename = innerfn
else:
debug("%s not in sources.tsv" % innerfn)
CaptureTime = iso8601(dt)
ExternalSource = args.extsrc or parse_pathname(
input_source).filename
SourceFilename = innerfn
ReceivedTime = iso8601(time.time())
InternalSource = args.intsrc or parse_pathname(
input_source).filename
already_received = metadb.check_already_received(
ExternalSource, SourceFilename)
xdid = ""
prev_xdid = "" # unshelved by default
existing_xdids = set(r.xdid for r in already_received)
if existing_xdids:
if len(existing_xdids) > 1:
warn(
'previously received this same file under multiple xdids:'
+ ' '.join(existing_xdids))
else:
prev_xdid = existing_xdids.pop()
debug('already received as %s' % prev_xdid)
# try each parser by extension
ext = parse_pathname(fn).ext.lower()
possible_parsers = parsers.get(ext, parsers[".puz"])
progress(fn)
if ext == ".xd":
outf.write_file(fn, contents.decode('utf-8'), dt)
elif not possible_parsers:
rejected = "no parser"
else:
rejected = ""
for parsefunc in possible_parsers:
try:
try:
xd = parsefunc(contents, fn)
except IncompletePuzzleParse as e:
error("%s %s" % (fn, e))
xd = e.xd
if not xd:
continue
xd.filename = replace_ext(strip_toplevel(fn),
".xd")
if not xd.get_header("Copyright"):
if args.copyright:
xd.set_header("Copyright", args.copyright)
catalog.deduce_set_seqnum(xd)
xdstr = xd.to_unicode()
mdtext = "|".join((ExternalSource, InternalSource,
SourceFilename))
xdid = prev_xdid or catalog.deduce_xdid(xd, mdtext)
path = catalog.get_shelf_path(
xd, args.pubid, mdtext)
outf.write_file(path + ".xd", xdstr, dt)
rejected = ""
break # stop after first successful parsing
except xdfile.NoShelfError as e:
error("could not shelve: %s" % str(e))
rejected += "[shelver] %s " % str(e)
except Exception as e:
error("%s could not convert [%s]: %s" %
(parsefunc.__name__, fn, str(e)))
rejected += "[%s] %s " % (parsefunc.__name__,
str(e))
# raise
if rejected:
error("could not convert: %s" % rejected)
# only add receipt if first time converting this source
if already_received:
debug("already received %s:%s" %
(ExternalSource, SourceFilename))
else:
receipts.append([
CaptureTime, ReceivedTime, ExternalSource,
InternalSource, SourceFilename, xdid
])
for r in receipts:
metadb.append_row('gxd/receipts', r)
except Exception as e:
error(str(e))
if args.debug:
raise
def main():
p = utils.args_parser(desc="annotate puzzle clues with earliest date used in the corpus")
p.add_argument("-a", "--all", default=False, help="analyze all puzzles, even those already in similar.tsv")
p.add_argument("-l", "--limit", default=100, help="limit amount of puzzles to be analyzed [default=100]")
args = get_args(parser=p)
outf = open_output()
num_processed = 0
prev_similar = metadb.read_rows("gxd/similar")
for fn, contents in find_files(*args.inputs, ext=".xd"):
progress(fn)
mainxd = xdfile(contents.decode("utf-8"), fn)
if mainxd.xdid() in prev_similar:
continue # skip reprocessing .xd that are already in similar.tsv
""" find similar grids (pct, xd) for the mainxd in the corpus.
Takes about 1 second per xd. sorted by pct.
"""
similar_grids = sorted(find_similar_to(mainxd, corpus(), min_pct=0.20), key=lambda x: x[0], reverse=True)
num_processed += 1
if num_processed > int(args.limit):
break
if similar_grids:
info("similar: " + " ".join(("%s=%s" % (xd2.xdid(), pct)) for pct, xd1, xd2 in similar_grids))
mainpubid = mainxd.publication_id()
maindate = mainxd.date()
# go over each clue/answer, find all other uses, other answers, other possibilities.
# these are added directly to similar.tsv
nstaleclues = 0
nstaleanswers = 0
ntotalclues = 0
for pos, mainclue, mainanswer in mainxd.iterclues():
progress(mainanswer)
poss_answers = []
pub_uses = {} # [pubid] -> set(ClueAnswer)
mainca = ClueAnswer(mainpubid, maindate, mainanswer, mainclue)
# find other uses of this clue, and other answers, in a single pass
for clueans in find_clue_variants(mainclue):
if clueans.answer != mainanswer:
poss_answers.append(clueans)
if clueans.answer == mainanswer:
if clueans.pubid in pub_uses:
otherpubs = pub_uses[clueans.pubid]
else:
otherpubs = set() # set of ClueAnswer
pub_uses[clueans.pubid] = otherpubs
otherpubs.add(clueans)
# bclues is all boiled clues for this particular answer: { [bc] -> #uses }
bclues = load_answers().get(mainanswer, [])
stale_answer = False
if bclues:
uses = []
for bc, nuses in bclues.items():
# then find all clues besides this one
clue_usages = [
ca for ca in load_clues().get(bc, []) if ca.answer == mainanswer and ca.date < maindate
]
if clue_usages:
stale_answer = True
if nuses > 1:
# only use one (the most recent) ClueAnswer per boiled clue
# but use the clue only (no xdid)
ca = sorted(clue_usages, key=lambda ca: ca.date or "z")[-1].clue
else:
ca = sorted(clue_usages, key=lambda ca: ca.date or "z")[-1]
uses.append((ca, nuses))
# summary row to similar.tsv
metadb.append_row(
"gxd/similar",
[
mainxd.xdid(), # xdid
int(100 * sum(pct / 100.0 for pct, xd1, xd2 in similar_grids)), # similar_grid_pct
nstaleclues, # reused_clues
nstaleanswers, # reused_answers
ntotalclues, # total_clues
" ".join(("%s=%s" % (xd2.xdid(), pct)) for pct, xd1, xd2 in similar_grids), # matches
],
)
def main():
args = utils.get_args('generate pub-years data')
outf = utils.open_output()
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
pubyears = {} # set()
for xd in xdfile.corpus():
puby = (xd.publication_id(), xd.year())
if puby not in pubyears:
pubyears[puby] = []
pubyears[puby].append(xd)
if pubyears:
metadb.delete_stats()
for puby, xdlist in sorted(pubyears.items()):
pubid, year = puby
npublic = 0
# TODO: SELECT FROM publications
nexisting = 0
# organize by day-of-week
byweekday = {}
byweekday_similar = {}
for w in weekdays:
byweekday[w] = []
byweekday_similar[w] = []
for xd in xdlist:
dow = dow_from_date(xd.get_header('Date'))
if dow: # Might be empty date or only a year
byweekday[dow].append(xd)
for r in metadb.xd_similar(pubid + str(year)):
if r.match_pct < 25:
continue
xd = xdfile.get_xd(r.xdid)
if xd:
dt = xd.get_header('Date')
if dt:
dow = dow_from_date(dt)
if dow: # Might be empty date or only a year
byweekday_similar[dow].append(r)
else:
debug("Date not set for: %s" % xd)
# tally stats
for weekday in weekdays:
copyrights = Counter() # [copyright_text] -> number of xd
editors = Counter() # [editor_name] -> number of xd
formats = Counter() # ["15x15 RS"] -> number of xd
# todo
nexisting = 0
nxd = len(byweekday[weekday])
public_xdids = [] # Empty for now
for xd in byweekday[weekday]:
xdid = xd.xdid()
if (year.isdigit()
and int(year) <= 1965) or xdid in public_xdids:
npublic += 1
editor = xd.get_header('Editor').strip()
if editor:
editors[editor] += 1
sizestr = xd.sizestr()
if sizestr:
formats[sizestr] += 1
copyright = xd.get_header('Copyright').strip()
if copyright:
copyrights[copyright] += 1
# debug("ME: %s MCPR: %s MF: %s" % (list(editors), list(copyrights), list(formats)))
def process_counter(count, comp_value):
# Process counter comparing with comp_value
if count:
item, num = count.most_common(1)[0]
if num != comp_value:
item += " (%s)" % num
else:
item = ''
return item
#
maineditor = process_counter(editors, nxd)
maincopyright = process_counter(copyrights, nxd)
mainformat = process_counter(formats, nxd)
reprints = 0
touchups = 0
redones = 0
copies = 0
themecopies = 0
for r in byweekday_similar[weekday]:
xd1 = xdfile.get_xd(r.xdid)
xd2 = xdfile.get_xd(r.match_xdid)
if xd1 is None:
info("%s: similar puzzle %s not in corpus" %
(r.match_xdid, r.xdid))
continue
if xd2 is None:
info("%s: similar puzzle %s not in corpus" %
(r.xdid, r.match_xdid))
continue
dt1 = xd1.get_header('Date')
dt2 = xd2.get_header('Date')
aut1 = xd1.get_header('Author').lower()
aut2 = xd2.get_header('Author').lower()
pct = int(r.match_pct)
if dt2 < dt1: # only capture the later one
##deduce_similarity_type
if diff_authors(aut1, aut2): # suspicious
if pct >= 50:
copies += 1
elif pct >= 30:
themecopies += 1
else:
if pct == 100:
reprints += 1
elif pct >= 50:
touchups += 1
elif pct >= 30:
themecopies += 1
metadb.append_row(
"pub/stats", (pubid, year, weekday, mainformat, maineditor,
maincopyright, nexisting, nxd, npublic, reprints,
touchups, redones, copies, themecopies))
def main():
p = utils.args_parser(
desc="annotate puzzle clues with earliest date used in the corpus")
p.add_argument(
'-a',
'--all',
default=False,
help='analyze all puzzles, even those already in similar.tsv')
p.add_argument('-l',
'--limit',
default=100,
help='limit amount of puzzles to be analyzed [default=100]')
args = get_args(parser=p)
outf = open_output()
num_processed = 0
prev_similar = metadb.read_rows('gxd/similar')
for fn, contents in find_files(*args.inputs, ext=".xd"):
progress(fn)
mainxd = xdfile(contents.decode('utf-8'), fn)
if mainxd.xdid() in prev_similar:
continue # skip reprocessing .xd that are already in similar.tsv
""" find similar grids (pct, xd) for the mainxd in the corpus.
Takes about 1 second per xd. sorted by pct.
"""
similar_grids = sorted(find_similar_to(mainxd, corpus(), min_pct=0.20),
key=lambda x: x[0],
reverse=True)
num_processed += 1
if num_processed > int(args.limit):
break
if similar_grids:
info("similar: " + " ".join(("%s=%s" % (xd2.xdid(), pct))
for pct, xd1, xd2 in similar_grids))
mainpubid = mainxd.publication_id()
maindate = mainxd.date()
# go over each clue/answer, find all other uses, other answers, other possibilities.
# these are added directly to similar.tsv
nstaleclues = 0
nstaleanswers = 0
ntotalclues = 0
for pos, mainclue, mainanswer in mainxd.iterclues():
progress(mainanswer)
poss_answers = []
pub_uses = {} # [pubid] -> set(ClueAnswer)
mainca = ClueAnswer(mainpubid, maindate, mainanswer, mainclue)
# find other uses of this clue, and other answers, in a single pass
for clueans in find_clue_variants(mainclue):
if clueans.answer != mainanswer:
poss_answers.append(clueans)
if clueans.answer == mainanswer:
if clueans.pubid in pub_uses:
otherpubs = pub_uses[clueans.pubid]
else:
otherpubs = set() # set of ClueAnswer
pub_uses[clueans.pubid] = otherpubs
otherpubs.add(clueans)
# bclues is all boiled clues for this particular answer: { [bc] -> #uses }
bclues = load_answers().get(mainanswer, [])
stale_answer = False
if bclues:
uses = []
for bc, nuses in bclues.items():
# then find all clues besides this one
clue_usages = [
ca for ca in load_clues().get(bc, [])
if ca.answer == mainanswer and ca.date < maindate
]
if clue_usages:
stale_answer = True
if nuses > 1:
# only use one (the most recent) ClueAnswer per boiled clue
# but use the clue only (no xdid)
ca = sorted(clue_usages,
key=lambda ca: ca.date or "z")[-1].clue
else:
ca = sorted(clue_usages,
key=lambda ca: ca.date or "z")[-1]
uses.append((ca, nuses))
# summary row to similar.tsv
metadb.append_row(
'gxd/similar',
[
mainxd.xdid(), # xdid
int(100 * sum(
pct / 100.0
for pct, xd1, xd2 in similar_grids)), # similar_grid_pct
nstaleclues, # reused_clues
nstaleanswers, # reused_answers
ntotalclues, # total_clues
" ".join(("%s=%s" % (xd2.xdid(), pct))
for pct, xd1, xd2 in similar_grids) # matches
])