def get_subfigure_f(db_file, image_dir, subfigures_dst):
conn = sqlite3.connect(db_file)
df = select_helper(conn, sql_select_empty_subfigurejsonfiles,
['pmcid', 'figure_name'])
insert_time = f'{datetime.datetime.now():%Y-%m-%d-%H-%M-%S}'
insert_subfigure_helper = DBHelper(conn, sql_insert_subfigure)
insert_subfigure_helper.start()
for pmcid, figure_name in tqdm.tqdm(zip(df['pmcid'], df['figure_name']),
total=len(df),
desc='Query new subfigures'):
figure_file = image_dir / generate_path(pmcid) / '{}_{}'.format(
pmcid, figure_name)
records = split_subfigure(figure_file)
records_to_insert = [(pmcid, figure_name, r.xtl, r.ytl, r.xbr, r.ybr,
'PMCFigureX', insert_time) for r in records]
insert_subfigure_helper.extend(records_to_insert)
insert_subfigure_helper.finish()
insert_subfigure_helper.summarize()
# get all subfigures
df = select_helper(conn, sql_select_subfigure,
['pmcid', 'figure_name', 'xtl', 'ytl', 'xbr', 'ybr'])
data = []
for i, row in tqdm.tqdm(df.iterrows(),
total=len(df),
desc='Get subfigures'):
pmcid = row['pmcid']
figure_name = row['figure_name']
xtl = row['xtl']
ytl = row['ytl']
xbr = row['xbr']
ybr = row['ybr']
figure_file = generate_path(pmcid) / '{}_{}'.format(pmcid, figure_name)
dst = figure_file.parent / f'{figure_file.stem}_{xtl}x{ytl}_{xbr}x{ybr}{figure_file.suffix}'
data.append({
'pmcid': pmcid,
'figure path': str(dst.as_posix()),
'xtl': xtl,
'ytl': ytl,
'xbr': xbr,
'ybr': ybr,
'type': 'subfigure'
})
df = pd.DataFrame(data)
df.to_csv(subfigures_dst, index=False)
conn.close()
def get_bioc_f(db_file, bioc_dir):
conn = sqlite3.connect(db_file)
df = select_helper(conn, sql_select_empty_bioc, ['pmcid', 'pmid'])
#
cnt = collections.Counter()
update_article_helper = DBHelper(conn, sql_update_articles)
update_article_helper.start()
for pmcid, pmid in tqdm.tqdm(zip(df['pmcid'], df['pmid']), total=len(df)):
cnt['total pmc'] += 1
dst_dir = bioc_dir / generate_path(pmcid)
dst = dst_dir / f'{pmcid}.xml'
if dst.exists():
update_article_helper.append((1, pmcid))
else:
try:
get_bioc(pmid, dst)
cnt['new bioc'] += 1
update_article_helper.append((1, pmcid))
except urllib.error.HTTPError:
update_article_helper.append((0, pmcid))
update_article_helper.finish()
conn.close()
for k, v in cnt.most_common():
print(k, ':', v)
def get_figures(db_file, image_dir):
conn = sqlite3.connect(db_file)
df = select_helper(conn, sql_get_empty_figures, ['pmcid', 'figure_name'])
cnt = collections.Counter()
update_figure_helper = DBHelper(conn, sql_update_figure_size)
update_figure_helper.start()
for pmcid, figure_name in tqdm.tqdm(zip(df['pmcid'], df['figure_name']),
total=len(df)):
local_file = image_dir / generate_path(pmcid) / '{}_{}'.format(
pmcid, figure_name)
if not local_file.exists():
try:
url = f'https://www.ncbi.nlm.nih.gov/pmc/articles/{pmcid}/bin/{figure_name}'
urllib.request.urlretrieve(url, local_file)
cnt['new figure'] += 1
except urllib.error.HTTPError:
cnt['Http error'] += 1
with open(local_file, 'w') as _:
pass
try:
im = Image.open(local_file)
update_figure_helper.append(
(im.width, im.height, pmcid, figure_name))
except:
cnt['Image error'] += 1
cnt['total figure'] += 1
update_figure_helper.finish()
conn.close()
ppprint.pprint_counter(cnt, percentage=False)
def move1(src_dir, dst_dir):
with os.scandir(src_dir) as it:
for entry in tqdm.tqdm(it):
src = entry.path
pmcid = Path(src).stem
parent_dir = dst_dir / generate_path(pmcid)
parent_dir.mkdir(parents=True, exist_ok=True)
dst = parent_dir / f'{pmcid}.xml'
shutil.move(src, dst)
def get_figure_url(db_file, bioc_dir):
conn = sqlite3.connect(db_file)
df = select_helper(conn, sql_select_new_bioc, ['pmcid'])
#
insert_time = f'{datetime.datetime.now():%Y-%m-%d-%H-%M-%S}'
insert_helper = DBHelper(conn, sql_insert_figure)
insert_helper.start()
for pmcid in tqdm.tqdm(df['pmcid'], total=len(df)):
biocfile = bioc_dir / generate_path(pmcid) / f'{pmcid}.xml'
figure_names = get_figure_link(biocfile)
insert_helper.extend(
set([(pmcid, figure_name, insert_time)
for figure_name in figure_names]))
insert_helper.finish()
conn.close()
def get_figure_text(src1, src2, dest, history_file, bioc_dir):
df1 = pd.read_csv(src1, dtype=str)
df2 = pd.read_csv(src2, dtype=str)
df = pd.concat([df1, df2], axis=0)
figures = create_figures(df, history_file=history_file)
docs = {} # type: Dict[str, bioc.BioCDocument]
for figure in figures:
pmcid = figure.pmcid
if pmcid not in docs:
src = bioc_dir / generate_path(pmcid) / f'{pmcid}.xml'
collection = bioc.load(open(src))
docs[pmcid] = collection.documents[0]
add_text(figure, docs[figure.pmcid])
with open(dest, 'w', encoding='utf8') as fp:
objs = [f.to_dict() for f in figures]
json.dump(objs, fp, indent=2)
def split_figure_f(db_file, image_dir, model_pathname, batch_size=16):
conn = sqlite3.connect(db_file)
df = select_helper(conn, sql_select_empty_subfigurejsonfiles,
['pmcid', 'figure_name'])
conn.close()
cnt = collections.Counter()
tf.compat.v1.disable_eager_execution()
separator = FigureSeparator(str(model_pathname))
with tf.compat.v1.Session(graph=separator.graph) as sess:
needs_to_split = []
def split_and_save():
srcs = [r[0] for r in needs_to_split]
dsts = [r[1] for r in needs_to_split]
results = separator.extract_batch(sess, srcs)
assert len(results) == len(srcs)
for dst, result in zip(dsts, results):
subfigures = result['sub_figures']
json.dump(subfigures, open(dst, 'w'))
for pmcid, figure_name in tqdm.tqdm(zip(df['pmcid'],
df['figure_name']),
total=len(df)):
src = image_dir / generate_path(pmcid) / '{}_{}'.format(
pmcid, figure_name)
dst = src.with_suffix('.json')
if not dst.exists():
needs_to_split.append((src, dst))
if len(needs_to_split) >= batch_size:
split_and_save()
needs_to_split = []
else:
if is_file_empty(src):
cnt['empty figure'] += 1
continue
if len(needs_to_split) > 0:
split_and_save()
for k, v in cnt.most_common():
print(k, ':', v)
def get_figure_f(db_file, image_dir, figures_dst):
conn = sqlite3.connect(db_file)
df = select_helper(conn, sql_select_figure,
['pmcid', 'figure_name', 'width', 'height'])
data = []
for i, row in tqdm.tqdm(df.iterrows(),
total=len(df),
desc='Get whole figures'):
pmcid = row['pmcid']
figure_name = row['figure_name']
figure_file = generate_path(pmcid) / '{}_{}'.format(pmcid, figure_name)
data.append({
'pmcid': pmcid,
'figure path': str(figure_file.as_posix()),
'xtl': 0,
'ytl': 0,
'xbr': row['width'],
'ybr': row['height'],
'type': 'figure'
})
df = pd.DataFrame(data)
df.to_csv(figures_dst, index=False)
conn.close()