# (3/4) téléchargement des fulltexts ---------------------------------
my_ids = cobj.cols['istex_id']
my_basenames = cobj.bnames
for the_shelf in ['PDF0', 'XMLN']:
the_api_type = cobj.origin(the_shelf)
the_ext = cobj.filext(the_shelf)
tgt_dir = cobj.shelf_path(the_shelf)
print("mkdir -p: %s" % tgt_dir)
mkdir(tgt_dir)
api.write_fulltexts_loop_interact(
my_ids, my_basenames,
tgt_dir = tgt_dir,
api_types = [the_api_type]
)
print("MAKE_SET: saved docs into CORPUS_HOME:%s" % cobj.name)
if debug > 0:
print(" (=> target dir:%s)" % tgt_dir)
# NB: il doit y avoir la même extension dans cobj.filext(the_shelf) que chez l'API
# ou alors api.write_fulltexts doit autoriser à changer (renommer) les extensions
cobj.assert_docs('PDF0')
cobj.assert_docs('XMLN')
# persistance du statut des 2 dossiers créés
cobj.save_shelves_status()
def make_set(corpus_name,
from_table=None,
size=None,
constraint=None,
debug=0):
"""
Initialisation d'un corpus basique et remplissage de ses fulltexts
3 façons de l'appeler :
- soit on fournit une table de métadonnées infos.tab (chemin fs)
- soit on fournit une taille (sampling directement avec l'API)
- soit on ne fournit rien et il fait un sampling de 10 docs
Métadonnées, rangées dans CORPUS_HOME/<corpus_name>/meta/
- basenames.ls
- infos.tab
Données: 3 formats, rangés dans CORPUS_HOME/<corpus_name>/data/
- .pdf,
- .xml (natif)
- et .tei.xml (pub2tei)
Position dans le système de fichier
cf sous lib/global_conf.ini
-> variable CORPUS_HOME
-> mise par défaut à ./corpora/
"""
# test de base avant de tout mettre dans le dossier
# (le seul dossier qu'on n'écrase jamais)
future_dir = path.join(MY_CORPUS_HOME, corpus_name)
if path.exists(future_dir):
print("ERR:'%s'\nLe nom '%s' est déjà pris dans le dossier des corpus." %
(future_dir, corpus_name), file=stderr)
exit(1)
# (1/4) echantillon initial (juste la table) -------------------------
# soit on a déjà une table
if from_table and size:
print("""ERR bako.make_set:
fournir au choix 'from_table' ou 'size', mais pas les 2.""",
file=stderr)
exit(1)
if from_table:
if path.exists(from_table):
fic = open(from_table)
my_tab = fic.readlines()
fic.close()
else:
print("ERR bako.make_set: je ne trouve pas le fichier '%s'" % from_table, file=stderr)
exit(1)
# sinon sampling
else:
if not size:
size = 10
if not constraint:
ok_corpora = CONF['workshop']['PREPROC_READY_CORPORA']
corpora_constraint = " OR ".join(['corpusName:'+corpus for corpus in ok_corpora.split(",")])
constraint = "qualityIndicators.refBibsNative:true AND (" + corpora_constraint +")"
if isinstance(size, int):
my_tab = sampler.full_run(
['-n', str(size),
'--outmode', 'tab',
'--with', constraint,
'-v'
]
)
else:
print("ERR bako.make_set: 'size' doit être un entier'%s'"
% from_table, file=stderr)
exit(1)
# (2/4) notre classe corpus ------------------------------------------
# Corpus
# initialisation
# - mode tab seul => fera un dossier meta/ et un data/ vide,
# - le corpus_type est mis en dur à 'gold' ce qui signale
# simplement qu'on ne change pas les étagères par défaut)
cobj = Corpus(corpus_name,
new_infos = my_tab,
new_home = MY_CORPUS_HOME,
verbose = (debug>0),
corpus_type='gold')
# (3/4) téléchargement des fulltexts ---------------------------------
my_ids = cobj.cols['istex_id']
my_basenames = cobj.bnames
for the_shelf in ['PDF0', 'XMLN']:
the_api_type = cobj.origin(the_shelf)
the_ext = cobj.filext(the_shelf)
tgt_dir = cobj.shelf_path(the_shelf)
print("mkdir -p: %s" % tgt_dir,file=stderr)
makedirs(tgt_dir)
api.write_fulltexts_loop_interact(
my_ids, my_basenames,
api_conf = CONF['istex-api'],
tgt_dir = tgt_dir,
api_types = [the_api_type]
)
print("MAKE_SET: saved docs into CORPUS_HOME:%s" % cobj.name)
if debug > 0:
print(" (=> target dir:%s)" % tgt_dir)
# NB: il doit y avoir la même extension dans cobj.filext(the_shelf) que chez l'API
# ou alors api.write_fulltexts doit autoriser à changer (renommer) les extensions
cobj.assert_docs('PDF0')
cobj.assert_docs('XMLN')
# persistance du statut des 2 dossiers créés
cobj.save_shelves_status()
# (4/4) conversion tei (type gold biblStruct) ------------------------
# copie en changeant les pointeurs dtd
print("***DTD LINKING***")
cobj.dtd_repair(debug_lvl = debug)
print("***XML => TEI.XML CONVERSION***")
# créera le dossier C-goldxmltei
cobj.pub2goldtei(debug_lvl = debug) # conversion
cobj.assert_docs('GTEI')
# persistence du statut du dossier créé
cobj.save_shelves_status()
# we return the new filled corpus for further work or display
return cobj
def full_run(arglist=None):
global LOG
global LISSAGE
# output lines for direct use or print to STDOUT if __main__
output_array = []
# cli arguments
args = my_parse_args(arglist)
# do we need to change smoothing ?
if args.smoothing_init and float(args.smoothing_init) > 0:
print("Setting initial smoothing to %.2f" % args.smoothing_init, file=stderr)
# global var change in main
LISSAGE = args.smoothing_init
# event log lines
LOG = ['INIT: sampling %i' % args.sample_size]
LOG.append('CRIT: fields(%s)' % ", ".join(args.criteria_list))
if args.with_constraint_query:
LOG.append('WITH: constraint query "%s"' % args.with_constraint_query)
run_counter = 0
# initial sampler run
got_ids_idx = sample(
args.sample_size,
args.criteria_list,
constraint_query = args.with_constraint_query,
verbose = args.verbose,
run_count = run_counter
)
run_counter += 1
# how much is there?
n_ids = len(got_ids_idx)
# info
print('-'*27 + " initial result : %i docs " % n_ids + '-'*27,
file=stderr)
LOG.append("XGOT: picked %i" % n_ids)
# check combopools status
insufficient_pool_flag = False
for sig in LOG:
if search("^LESS:", sig):
insufficient_pool_flag = True
break
# --------- a posteriori corrections -------------
#
# the initial quotas can take neither the "with_constraint arg"
# nor "multiple choice fields" into account (unless use N_reponse?)
# for that reason at this point in the process we may have more or
# less than the requested sample_size
# IF not enough => new sample run with lighter criteria
if n_ids < args.sample_size:
actual_criteria = args.criteria_list
# keep trying...
while (n_ids < args.sample_size and run_counter < MAX_RUNS):
# => over "delta" (missing docs)
remainder = args.sample_size - n_ids
LOG.append("REDO: re-pioche sur %i docs" % remainder)
# => with more help to small categories
LISSAGE += 0.2
LOG.append("SMOO: smoothing up to %.02f" % LISSAGE)
# => and with less criteria if necessary
# (if criteria pool insufficient under some constraints, we
# do need to relax at least one criterion, but which one?)
if len(actual_criteria) > 1 and insufficient_pool_flag:
# simplify criteria by removing the last one
new_criteria = actual_criteria[0:-1]
LOG.append("RLAX: abandon équilibrage champ '%s'" %
actual_criteria[-1])
# reset flag (£TODO recalculate after run ?)
insufficient_pool_flag = False
else:
new_criteria = actual_criteria
# -------- RE-RUN ---------
previous_ids = got_ids_idx
got_ids_idx = sample(
remainder,
new_criteria,
constraint_query = args.with_constraint_query,
index = previous_ids,
verbose = args.verbose
)
# recount
apport = len(got_ids_idx) - n_ids
# update
n_ids += apport
run_counter += 1
# warn
LOG.append("XGOT: picked %i" % apport)
print('-'*22 + " resultat après run %i: %i documents "
% (run_counter, n_ids) + '-'*22, file=stderr)
# IF overflow => random pruning
if n_ids > args.sample_size:
deck = [did for did in got_ids_idx.keys()]
# random removal of excess documents
shuffle(deck)
nd = n_ids - args.sample_size
sacrificed = deck[0:nd]
for did in sacrificed:
del got_ids_idx[did]
LOG.append("XDEL: sacrificing %i random docs" % nd)
# last recount
n_ids = len(got_ids_idx)
print('-'*29 +" final result: %i docs "%n_ids+'-'*29, file=stderr)
# -------------- OUTPUT --------------------------------------------
# ***(ids)***
if args.out_type == 'ids':
for did, info in sorted(got_ids_idx.items(), key=lambda x: x[1]['_q']):
output_array.append("%s" % did)
# ***(tab)***
elif args.out_type == 'tab':
# header line
# £TODO STD_MAP
output_array.append("\t".join(['istex_id', 'corpus', 'pub_year', 'pub_period', 'pdfver', 'pdfwc','bibnat',
'author_1','lang','doctype_1','cat_sci', 'title']))
# contents
for did, info in sorted(got_ids_idx.items(), key=lambda x: x[1]['_q']):
# provenance: sample() => boucle par hits (l.500 ++)
# print("INFO----------",info, file=stderr)
# exit()
period = year_to_range(info['yr'])
output_array.append("\t".join([ did,
info['co'],
info['yr'],
period,
info['ver'],
str(info['wcp']),
str(info['bibnat']),
info['au'],
info['lg'],
info['typ'],
info['cat'],
info['ti'],
#~ info['_q']
]
)
)
# ***(docs)***
# no output lines but writes a dir
elif args.out_type == 'docs':
my_dir = path.join(getcwd(),my_name)
mkdir(my_dir)
# two "parallel" lists
ids = list(got_ids_idx.keys())
basenames = [std_filename(one_id, got_ids_idx[one_id]) for one_id in ids]
# loop with interactive authentification prompt if needed
api.write_fulltexts_loop_interact(
ids, basenames,
tgt_dir=my_dir,
api_types=['metadata/xml',
'fulltext/pdf',
'fulltext/tei']
)
LOG.append("SAVE: saved docs in %s/" % my_dir)
if args.log:
# separate logging lines
logfile = open(my_name+'.log', 'w')
for lline in LOG:
print(lline, file=logfile)
logfile.close()
return output_array
# (3/4) téléchargement des fulltexts ---------------------------------
my_ids = cobj.cols['istex_id']
my_basenames = cobj.bnames
for the_shelf in ['PDF0', 'XMLN']:
the_api_type = cobj.origin(the_shelf)
the_ext = cobj.filext(the_shelf)
tgt_dir = cobj.shelf_path(the_shelf)
print("mkdir -p: %s" % tgt_dir)
mkdir(tgt_dir)
api.write_fulltexts_loop_interact(my_ids,
my_basenames,
tgt_dir=tgt_dir,
api_types=[the_api_type])
print("MAKE_SET: saved docs into CORPUS_HOME:%s" % cobj.name)
if debug > 0:
print(" (=> target dir:%s)" % tgt_dir)
# NB: il doit y avoir la même extension dans cobj.filext(the_shelf) que chez l'API
# ou alors api.write_fulltexts doit autoriser à changer (renommer) les extensions
cobj.assert_docs('PDF0')
cobj.assert_docs('XMLN')
# persistance du statut des 2 dossiers créés
cobj.save_shelves_status()
# (4/4) conversion tei (type gold biblStruct) ------------------------
def full_run(arglist=None):
global LOG
global LISSAGE
# output lines for direct use or print to STDOUT if __main__
output_array = []
# cli arguments
args = my_parse_args(arglist)
# do we need to change smoothing ?
if args.smoothing_init and float(args.smoothing_init) > 0:
print("Setting initial smoothing to %.2f" % args.smoothing_init,
file=stderr)
# global var change in main
LISSAGE = args.smoothing_init
# event log lines
LOG = ['INIT: sampling %i' % args.sample_size]
LOG.append('CRIT: fields(%s)' % ", ".join(args.criteria_list))
if args.with_constraint_query:
LOG.append('WITH: constraint query "%s"' % args.with_constraint_query)
run_counter = 0
# initial sampler run
got_ids_idx = sample(args.sample_size,
args.criteria_list,
constraint_query=args.with_constraint_query,
verbose=args.verbose,
run_count=run_counter)
run_counter += 1
# how much is there?
n_ids = len(got_ids_idx)
# info
print('-' * 27 + " initial result : %i docs " % n_ids + '-' * 27,
file=stderr)
LOG.append("XGOT: picked %i" % n_ids)
# check combopools status
insufficient_pool_flag = False
for sig in LOG:
if search("^LESS:", sig):
insufficient_pool_flag = True
break
# --------- a posteriori corrections -------------
#
# the initial quotas can take neither the "with_constraint arg"
# nor "multiple choice fields" into account (unless use N_reponse?)
# for that reason at this point in the process we may have more or
# less than the requested sample_size
# IF not enough => new sample run with lighter criteria
if n_ids < args.sample_size:
actual_criteria = args.criteria_list
# keep trying...
while (n_ids < args.sample_size and run_counter < MAX_RUNS):
# => over "delta" (missing docs)
remainder = args.sample_size - n_ids
LOG.append("REDO: re-pioche sur %i docs" % remainder)
# => with more help to small categories
LISSAGE += 0.2
LOG.append("SMOO: smoothing up to %.02f" % LISSAGE)
# => and with less criteria if necessary
# (if criteria pool insufficient under some constraints, we
# do need to relax at least one criterion, but which one?)
if len(actual_criteria) > 1 and insufficient_pool_flag:
# simplify criteria by removing the last one
new_criteria = actual_criteria[0:-1]
LOG.append("RLAX: abandon équilibrage champ '%s'" %
actual_criteria[-1])
# reset flag (£TODO recalculate after run ?)
insufficient_pool_flag = False
else:
new_criteria = actual_criteria
# -------- RE-RUN ---------
previous_ids = got_ids_idx
got_ids_idx = sample(remainder,
new_criteria,
constraint_query=args.with_constraint_query,
index=previous_ids,
verbose=args.verbose)
# recount
apport = len(got_ids_idx) - n_ids
# update
n_ids += apport
run_counter += 1
# warn
LOG.append("XGOT: picked %i" % apport)
print('-' * 22 + " resultat après run %i: %i documents " %
(run_counter, n_ids) + '-' * 22,
file=stderr)
# IF overflow => random pruning
if n_ids > args.sample_size:
deck = [did for did in got_ids_idx.keys()]
# random removal of excess documents
shuffle(deck)
nd = n_ids - args.sample_size
sacrificed = deck[0:nd]
for did in sacrificed:
del got_ids_idx[did]
LOG.append("XDEL: sacrificing %i random docs" % nd)
# last recount
n_ids = len(got_ids_idx)
print('-' * 29 + " final result: %i docs " % n_ids + '-' * 29, file=stderr)
# -------------- OUTPUT --------------------------------------------
# ***(ids)***
if args.out_type == 'ids':
for did, info in sorted(got_ids_idx.items(), key=lambda x: x[1]['_q']):
output_array.append("%s" % did)
# ***(tab)***
elif args.out_type == 'tab':
# header line
# £TODO STD_MAP
output_array.append("\t".join([
'istex_id', 'corpus', 'pub_year', 'pub_period', 'pdfver', 'pdfwc',
'author_1', 'lang', 'doctype_1', 'cat_sci', 'title'
]))
# contents
for did, info in sorted(got_ids_idx.items(), key=lambda x: x[1]['_q']):
# provenance: sample() => boucle par hits (l.500 ++)
# print("INFO----------",info, file=stderr)
# exit()
period = year_to_range(info['yr'])
output_array.append("\t".join([
did,
info['co'],
info['yr'],
period,
info['ver'],
str(info['wcp']),
info['au'],
info['lg'],
info['typ'],
info['cat'],
info['ti'],
#~ info['_q']
]))
# ***(docs)***
# no output lines but writes a dir
elif args.out_type == 'docs':
my_dir = path.join(getcwd(), my_name)
mkdir(my_dir)
# two "parallel" lists
ids = list(got_ids_idx.keys())
basenames = [
std_filename(one_id, got_ids_idx[one_id]) for one_id in ids
]
# loop with interactive authentification prompt if needed
api.write_fulltexts_loop_interact(
ids,
basenames,
tgt_dir=my_dir,
api_types=['metadata/xml', 'fulltext/pdf'])
LOG.append("SAVE: saved docs in %s/" % my_dir)
return (output_array, LOG)
