def parse_folder(path):
"""
Parses all .bib files in given folder.
Returns a tuple (parsed_iten, search_index) containing all items found
"""
if not os.path.isdir(path):
raise Exception("Path to folder expected")
parsed_items = []
files = utils.files_in_folder(path, "*.bib")
executor = concurrent.futures.ProcessPoolExecutor(max_workers=multiprocessing.cpu_count())
futures = [
executor.submit(
BibParser()._parse_file,
os.path.join(path, filename)
)
for filename in files
]
for future in futures:
parsed_items += future.result()
executor.shutdown()
parsed_items = list(sorted(
parsed_items,
key=BibItem.key_to_key_func(const.DEFAULT_ORDER_BY)
))
item_index = search_index.Index(parsed_items)
fin_ctx = FinalizingContext(item_index)
for item in parsed_items:
item.finalize_item_set(fin_ctx)
item_index.update(parsed_items)
return (parsed_items, item_index)
def generate_output_template():
output_images = files_in_folder("./output/")
test_images = files_in_folder("./dataset/test/raw/")
true_masks_images = files_in_folder("./dataset/test/segmentation_mask/")
template_predictions = list()
for i in range(len(output_images)):
template_predictions.append(PredictionResult(test_images[i], output_images[i], true_masks_images[i]))
loader = FileSystemLoader("./templates")
env = Environment(loader=loader)
template = env.get_template('index.html')
template_output = template.render(predictions=template_predictions)
reports_path = "./reports"
if not os.path.exists(reports_path):
os.makedirs(reports_path)
text_file = open(f"{reports_path}/index.html", "w")
text_file.write(template_output)
text_file.close()
def parse_folder(self, path: str) -> [BibItem]:
"""
Parses all .bib files in given folder.
Returns list containing all items found
"""
if not os.path.isdir(path):
raise Exception("Path to folder expected")
parsed_items = []
files = utils.files_in_folder(path, "*.bib")
for filename in files:
parsed_items += self.parse_file(os.path.join(path, filename))
return parsed_items
def parse_folder(self, path):
"""
Parses all .bib files in given folder.
Returns list containing all items found
"""
if not os.path.isdir(path):
raise Exception("Path to folder expected")
parsed_items = []
files = utils.files_in_folder(path, "*.bib")
for filename in files:
parsed_items += self.parse_file(os.path.join(path, filename))
return parsed_items
def main():
if not os.path.isdir(config.www.elibrary_dir):
print("root folder '{elibrary_dir}' is inaccessible".format(
elibrary_dir=config.www.elibrary_dir
))
sys.exit(1)
#filename in database is relative, but begins from /
file_modifier = lambda file_, root=config.www.elibrary_dir: os.path.join(root, file_[1:])
filenames = set(map(file_modifier, item_index["filename"].keys()))
files = utils.files_in_folder(config.www.elibrary_dir, "*.pdf", excludes=EXCLUDED_FOLDERS)
files_filter = lambda file_: file_ not in filenames
files = list(filter(files_filter, files))
for file in files:
print(file)
def main():
if not os.path.isdir(config.www.elibrary_dir):
print("root folder '{elibrary_dir}' is inaccessible".format(
elibrary_dir=config.www.elibrary_dir))
sys.exit(1)
#filename in database is relative, but begins from /
file_modifier = lambda file_, root=config.www.elibrary_dir: os.path.join(
root, file_[1:])
filenames = set(map(file_modifier, item_index["filename"].keys()))
files = utils.files_in_folder(config.www.elibrary_dir,
"*.pdf",
excludes=EXCLUDED_FOLDERS)
files_filter = lambda file_: file_ not in filenames
files = list(filter(files_filter, files))
for file in files:
print(file)
def update_graphs(qrels, run1, run2, metric, top):
global qrels_dict
df_a = trec_eval(metric, "qrels/" + qrels, run1)
df_b = trec_eval(metric, "qrels/" + qrels, run2)
df = pd.concat([df_a, df_b], axis=1)
df = df.sort_values(run1, ascending=False)
# ranking
ranking_1 = read_run("runs/" + run1)
ranking_2 = read_run("runs/" + run2)
df_dict_a = {}
for topic in ranking_1.keys():
df_dict_a[topic] = pd.DataFrame(ranking_1[topic], columns=["docid"])
df_dict_a[topic][run1] = df_dict_a[topic]["docid"].apply(
lambda x: find_relevance(x, qrels_dict[topic]))
df_dict_b = {}
for topic in ranking_2.keys():
df_dict_b[topic] = pd.DataFrame(ranking_2[topic], columns=["docid"])
df_dict_b[topic][run2] = df_dict_b[topic]["docid"].apply(
lambda x: find_relevance(x, qrels_dict[topic]))
# merge dataframes
df_dict = {}
df_dict[top] = df_dict_a[top].merge(df_dict_b[top],
left_index=True,
right_index=True)
# set n to 5 if @5 or 10 if @10:
total = 1
n = int("".join(filter(str.isdigit, metric)))
block_size = total / n
# ndcg plot
fig_a = df.plot.bar(barmode="group")
fig_a.update_layout(
plot_bgcolor="white",
yaxis_title=pretty_metric[metric],
xaxis_title="Topic",
title=(f"{pretty_metric[metric]}: {df_a[run1].mean():.4f}"
f"({df_a[run1].median():.4f}) & {df_b[run2].mean():.4f}"
f" ({df_b[run2].median():.4f})"),
title_x=0.5,
)
# ranking annotations docids (should be put in separate function)
y1 = df_dict[top][run1].values
y2 = df_dict[top][run2].values
xcoord = df_dict[top].iloc[0:n].index # [0,1,2]
annotations1 = [
dict(
x=8, # xi-0.2,
y=xi - (total / 5) + (5 / 30),
text=df_dict[top]["docid_x"].iloc[xi],
xanchor="auto",
yanchor="bottom",
showarrow=False,
font={
"size": block_size * 75,
"color": "grey"
},
) for xi, yi in zip(xcoord, y1)
]
annotations2 = [
dict(
x=8,
y=xi + (total / 5) + (5 / 30),
text=mark_new_text(
df_dict[top]["docid_y"].iloc[xi],
df_dict[top]["docid_x"].iloc[0:n],
),
xanchor="auto",
yanchor="bottom",
showarrow=False,
font={
"size":
block_size * 75,
"color":
mark_new(df_dict[top]["docid_y"].iloc[xi],
df_dict[top]["docid_x"].iloc[0:n]),
},
) for xi, yi in zip(xcoord, y2)
]
annotations = annotations1 + annotations2
# ranking plot
fig_b = df_dict[top].iloc[0:n].plot.barh(x=[run1, run2], barmode="group")
fig_b.update_layout(
plot_bgcolor="white",
yaxis={
"tickmode": "array",
"tickvals": np.arange(n),
"ticktext": np.arange(1, n + 1),
"autorange": "reversed",
},
title=
f"{pretty_metric[metric]}: {df_a[run1].loc[top]} & {df_b[run2].loc[top]}",
title_x=0.5,
yaxis_title="Rank",
xaxis={
"range": [0, 16],
"fixedrange": True
},
xaxis_title="Relevance",
annotations=annotations,
)
df_all = pd.DataFrame()
for file in files_in_folder("runs"):
try:
df_tmp = trec_eval(metric, "qrels/" + qrels, file)
df_all = pd.concat([df_all, df_tmp], axis=1)
except Exception as e:
logging.info(f"{file} is not a TREC run. --> {e}")
continue
medians = df_all.median().sort_values()
medians_desc = df_all.median().sort_values(ascending=False)
traces = []
for run_name, run_data in df_all[medians.index].iteritems():
color = "rgb(0, 0, 0)"
if run_name == run1:
color = "rgb(99, 110, 250)"
if run_name == run2:
color = "rgb(239, 85, 59)"
traces.append(
go.Box(
y=run_data,
name=run_name,
boxpoints="all",
jitter=0.5,
whiskerwidth=0.2,
marker=dict(size=2, color=color),
line=dict(width=1),
))
fig_box = go.Figure(data=traces)
fig_box.update_layout(
plot_bgcolor="white",
title="Overview",
title_x=0.5,
yaxis_title=pretty_metric[metric],
showlegend=False,
)
headerColor = "grey"
fig_table = go.Figure(data=[
go.Table(
header=dict(
values=[
"<b>run</b>",
f"<b>percentage new in top {n} (relative to base)</b>",
f"<b>{pretty_metric[metric]} (mean : median)</b>",
"<b>p-value (H0: Equal average with base)</b>",
],
line_color="darkslategray",
fill_color=headerColor,
align=["left", "center"],
font=dict(color="white", size=12),
),
cells=dict(
values=[
df_all[medians_desc.index].columns,
[
"{:.2f}%".format(
new_percentage(ranking_1, read_run("runs/" +
file), n))
for file in df_all[medians_desc.index].columns
],
[(f"{round(np.mean(df_all[run]), 4):.4f} : "
f"{round(np.median(df_all[run]), 4):.4f}")
for run in df_all[medians_desc.index].columns],
[
f"{ttest_rel(df_a, df_all[col])[1][0]:.4f}"
if col != run1 else "X"
for col in df_all[medians_desc.index].columns
],
],
line_color="darkslategray",
fill_color=[[
mark_current_runs(run1, run2, run)
for run in df_all[medians_desc.index].columns
]],
align=["left", "center"],
font=dict(color="darkslategray", size=11),
),
)
])
fig_table.update_layout(
plot_bgcolor="white",
title="Table",
title_x=0.5,
yaxis_title=pretty_metric[metric],
)
return fig_a, fig_b, fig_box, fig_table
style={
"display": "flex",
"justify-content": "center"
},
),
html.Div(
[
html.Div(
[
html.P("base"),
dcc.Dropdown(
id="dropdown-run-1",
options=[{
"label": file,
"value": file
} for file in files_in_folder("runs")],
value=files_in_folder("runs")[0],
style={"width": 180},
),
],
style=style_block,
),
html.Div(
[
html.P("alternative"),
dcc.Dropdown(
id="dropdown-run-2",
options=[{
"label": file,
"value": file
} for file in files_in_folder("runs")],
def main(
max_count=("c", 100, "Maximum count of filenames to display"),
root=("r", "", "E-library root")
):
if (len(root) == 0) or (not os.path.isdir(root)):
print("Root folder is inaccessible")
sys.exit(1)
root = os.path.abspath(root)
#filename in database is relative, but begins from /
file_modifier = lambda file_, root=root: os.path.join(root, file_[1:])
filenames = set(map(file_modifier, item_index["filename"].keys()))
files = utils.files_in_folder(root, "*.pdf", excludes=EXCLUDED_FOLDERS)
files_filter = lambda file_: file_ not in filenames
files = list(filter(files_filter, files))
print("Going to process {0} items".format(len(items)))
print("Going to process {0} files".format(len(files)))
output_count = 0
output_dict = dict()
for file_ in files:
relpath = "/" + os.path.relpath(file_, root)
metadata = utils.extract_metadata_from_file(file_)
item_search = utils.create_search_from_metadata(metadata)
found_items = list(filter(item_search, items))
found_count = len(found_items)
if found_count == 0:
print("Nothing found for file '{relpath}'".format(
relpath=relpath,
))
elif found_count == 1:
item = found_items[0]
if item in output_dict:
output_dict[item].add(relpath)
else:
output_dict[item] = set([relpath])
else:
source_getter = lambda item: item.source()
print("Found multiple items for '{relpath}':\n\t{sources}".format(
sources=list(map(source_getter, found_items)),
relpath=relpath
))
output_count += 1
if output_count >= max_count:
print("Reached maxcount. Exiting")
break
sort_key = lambda pair: pair[0].source()
for item, paths in sorted(output_dict.items(), key=sort_key):
print("Filename for {id} ({source}):".format(
id=item.id(),
source=item.source(),
))
print("filename = {{{relpath}}}".format(
relpath=" {0} ".format(config.parser.list_sep).join(sorted(paths))
))
sort_key = lambda item: item.source()
if len(items) < max_count:
for item in sorted(items, key=sort_key):
print("Item isn't digitized: {id} ({source})".format(
id=item.id(),
source=item.source()))
import utils
utils.files_in_folder('.', 'files.csv')
