🐛 Only strip enclosing delimiters that enclose the whole value - #604
Merged
Conversation
MiWeiss
force-pushed
the
fix/enclosing-concatenation
branch
3 times, most recently
from
September 2, 2026 20:00
94d484d to
f313c6f
Compare
`_strip_enclosing` assumed a leading `{`/`"` is closed by the trailing one,
so concatenation expressions like `pages = {intro} # {outro}` or
`title = "x" # "y"` were stripped to corrupted, non-concatenation values and
misclassified as enclosed (and thus re-wrapped on write). A lone `"` was also
stripped to the empty string.
Brace stripping now tracks brace depth (ignoring backslash-escaped braces) and
only strips when the brace opened at index 0 is the one closed at the last
index; quote stripping only strips when no unescaped `"` occurs at brace-depth
zero in between. Everything else falls through to `no-enclosing`, which - as
the middleware docstring already demanded - makes such values round-trip
verbatim.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
MiWeiss
force-pushed
the
fix/enclosing-concatenation
branch
2 times, most recently
from
September 2, 2026 20:38
192ab12 to
5aa75ef
Compare
The `no-enclosing` demand and the delimiters it was derived from live in two
separate places: `{Foo} # {, Bar}` keeps its braces inside `field.value`, while
the demand sits on the field. The value-transforming middlewares deliberately
carry the demand across a value rewrite, so once such a middleware removes the
braces the demand is stale and the writer emits unparseable bibtex:
author = {Doe, John} and {Roe, Jane} -> author = Doe, John and Roe, Jane
Rather than chase every middleware, `AddEnclosingMiddleware` now falls back to
the default enclosing whenever a `no-enclosing` value could not be read back in
one piece, i.e. when it has an unbalanced brace or a comma, equals sign or
newline outside braces and quotes.
The two `_is_enclosed_in_*` helpers now scan delimiters with a regex instead of
every character. This aligns their escaping with the splitter's mark regex,
where a delimiter is escaped iff directly preceded by a backslash (so `{\\}a}`
is stripped again), and removes the parse-time cost of the per-character scan:
3000 entries with a 1.2 KB abstract go back from 0.220 s to 0.105 s, against
0.104 s before this branch.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
MiWeiss
added a commit
that referenced
this pull request
Sep 3, 2026
`_is_enclosed_in_quotes` (added in #604) scanned every quoted value with a regex, making the default parse stack ~4x slower on files with quoted values. Skip the scan when the value contains no inner quote, and avoid the slice copy in the corresponding brace fast path. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
_strip_enclosingonly testsstartswith/endswith, so any value built from several delimited parts is corrupted, then classified as enclosed and re-wrapped as a literal.This contradicts the middleware's own docstring on concatenations.
Fix: strip only a pair enclosing the whole value. Brace depth, escape-aware; for quotes, no unescaped
"at depth 0 inside. Everything else falls through tono-enclosingand is written verbatim.Tests: 27 cases, including shapes that must still strip (
{a {b} c},{{nested}},{},"", escaped braces). Suite 2603 passed, from 2576.🤖 Generated with Claude Code