Files
ma-thesis/src/thesis/clean_bibliography.sh

112 lines
3.8 KiB
Bash
Executable File

sed -i "s/Świerkowska/{\\\\'S}wierkowska/" bibliography.bib
sed -i "s/Héctor/H{\\\\'e}ctor/" bibliography.bib
sed -i "s/Bombín/Bomb{\\\\'i}n/" bibliography.bib
sed -i "s/Zémor/Z{\\\\'e}mor/" bibliography.bib
sed -Ezi "s/\s(abstract|note|urldate|url|keywords|file) = \{[^}]*(\{[^}]*\}[^}]*)*\},?\n//g" bibliography.bib
# Normalize arXiv-only entries to @misc with howpublished = {arXiv:<id>}.
# Detection: doi matches 10.48550/arXiv.<id>. The IEEEtranSA .bst's @article
# handler needs a journal field (which preprints lack) and ignores publisher,
# so for arXiv preprints we coerce the type to @misc and add howpublished
# (the field the .bst actually prints for @misc).
python3 - <<'PY'
import re
path = "bibliography.bib"
with open(path) as f:
text = f.read()
doi_re = re.compile(r"doi\s*=\s*\{10\.48550/arXiv\.([^}]+)\}")
type_re = re.compile(r"^@([A-Za-z]+)\{", re.MULTILINE)
howpublished_re = re.compile(r"^\s*howpublished\s*=\s*\{", re.MULTILINE)
title_field_re = re.compile(r"\b(title|booktitle)\s*=\s*\{", re.IGNORECASE)
inner_brace_re = re.compile(r"\{([A-Za-z0-9]+)\}")
# Split into entries by scanning for top-level "@type{...}" blocks. We walk
# brace depth so that the closing "}" of the entry is matched correctly even
# if internal fields contain braces.
def split_entries(s):
out, i, n = [], 0, len(s)
while i < n:
m = type_re.search(s, i)
if not m:
out.append(("text", s[i:]))
break
if m.start() > i:
out.append(("text", s[i:m.start()]))
depth, j = 0, m.start()
while j < n:
c = s[j]
if c == "{":
depth += 1
elif c == "}":
depth -= 1
if depth == 0:
j += 1
break
j += 1
out.append(("entry", s[m.start():j]))
i = j
return out
def normalize_arxiv(entry):
doi_m = doi_re.search(entry)
if not doi_m:
return entry
arxiv_id = doi_m.group(1)
entry = type_re.sub("@misc{", entry, count=1)
if not howpublished_re.search(entry):
# insert howpublished as the last field, before the entry-closing "}"
entry = re.sub(
r"(,?)(\s*)\}\s*$",
lambda m: ("," if m.group(1) != "," else m.group(1))
+ m.group(2) + "\thowpublished = {arXiv:" + arxiv_id + "},\n}",
entry,
count=1,
)
return entry
# Strip protective braces around words inside title/booktitle values.
# BibTeX uses "{Word}" inside titles to preserve case against the bibliography
# style's title-casing rules. We keep that protection only when every character
# inside the braces is non-lowercase (e.g. acronyms like {NASA}); for ordinary
# words like {Quantum} we drop the braces so the style's casing applies.
def strip_title_braces(entry):
out, i, n = [], 0, len(entry)
while True:
m = title_field_re.search(entry, i)
if not m:
out.append(entry[i:])
break
out.append(entry[i:m.end()])
depth, j = 1, m.end()
while j < n and depth > 0:
c = entry[j]
if c == "{":
depth += 1
elif c == "}":
depth -= 1
if depth == 0:
break
j += 1
value = entry[m.end():j]
cleaned = inner_brace_re.sub(
lambda mm: mm.group(1) if any(c.islower() for c in mm.group(1)) else mm.group(0),
value,
)
out.append(cleaned)
if j < n:
out.append(entry[j])
i = j + 1
return "".join(out)
def transform(entry):
return strip_title_braces(normalize_arxiv(entry))
parts = split_entries(text)
new_text = "".join(transform(p) if kind == "entry" else p for kind, p in parts)
with open(path, "w") as f:
f.write(new_text)
PY