format: normalize quote style across single, double, and backtick strings - #164
format: normalize quote style across single, double, and backtick strings#164claude[bot] wants to merge 4 commits into
Conversation
Convert string literals to a preferred quote style: double quotes by default, single quotes when the content contains a double quote, and backticks when it contains both a double and a single quote. Applies to rval strings (via the quoted_string choke point) and promisers.
c6a5d1b to
e2d6aa2
Compare
| Backtick strings are taken literally - CFEngine processes no escapes | ||
| inside them. Single- and double-quoted strings recognize only the escapes |
There was a problem hiding this comment.
This is not true, backtick strings are treated the same, and escapes are processed.
| if nxt == "\n": # line continuation: drop both characters | ||
| i += 2 | ||
| continue |
There was a problem hiding this comment.
This behavior was not requested and is not covered by the test.
| if c == "\\" and i + 1 < len(inner): | ||
| nxt = inner[i + 1] | ||
| if nxt in ("\\", '"', "'"): | ||
| out.append(nxt) | ||
| i += 2 | ||
| continue |
There was a problem hiding this comment.
The formatting test does not have \\ (or \\\\), should cover that too.
| if ( | ||
| len(literal) < 2 | ||
| or literal[0] != literal[-1] | ||
| or literal[0] not in ("'", '"', "`") | ||
| ): | ||
| return literal |
There was a problem hiding this comment.
Should this be an assert? I don't see when this would be the case?
| if target == "`" and "`" in content: | ||
| # Needs all three quote styles at once; a backtick string cannot | ||
| # contain a backtick, so leave the literal as written. | ||
| return literal |
There was a problem hiding this comment.
This is also not covered in the test.
Address review on PR #164: - Backtick strings process the SAME escapes as single/double quotes (\\, \", \'); a backtick string is not literal. Its only special property is that the delimiter itself cannot be escaped, so a literal backtick can never appear inside one. Unified _decode_literal (drops its delim arg) and _encode_literal (now escapes backslashes for backticks too) accordingly. - Removed the unrequested backslash-newline line-continuation handling from _decode_literal. - Turned the literal-shape guard in _normalize_quotes into an assert, since callers always pass a real quoted_string literal. - Expanded the 012_quotes golden fixture with backslash cases (l/m/n/p) and an all-three-quotes case (o) to cover the unified escape behavior and the backtick encode branch. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U4hEZuqiuEFH2zy8wWGwyD
Replaced the _decode_literal/_encode_literal helpers with a single-pass, escape-aware re-quote (_scan_quotes + _requote) so converting between quote styles passes through escaped backslashes, line continuations, and other backslash sequences unchanged, and only re-escapes the target delimiter quote character. The old decode/encode round-trip would mangle a single-quoted line-continuation string into a doubled backslash. Fixed fixture case q to convert a single-quoted line-continuation string to double quotes with the continuation preserved verbatim. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U4hEZuqiuEFH2zy8wWGwyD
nickanderson
left a comment
There was a problem hiding this comment.
I agree with the suggested expectations I just think this should require opting in, not done by default with format. Something link could surface. But then I or someone else will also ask for a way to suppress the warning with some kind of comment.
|
Marking this PR as stale due to inactivity; it will be closed in 7 days. |
|
I will postpone this a bit. I think it should be done by default in
However I believe we have quite a few tests which test different types of quotes, and in those cases we do want to format them, but not change the quotes. So we'll have to figure out what to do there. Closing for now. |
Requested by Ole Herman · Slack thread
Before / After
cfengine formatnow normalizes string-literal quote style across double, single, and backtick quotes. It prefers double quotes; falls back to single quotes when the string contains a double-quote character; and uses backticks only when the string contains both a double and a single quote.Examples:
'hello'becomes"hello"`hello`becomes"hello"'say "hi"'stays single-quoted (it contains a double quote but no single quote)`say "hi"`becomes'say "hi"''he said "hi" it\'s') becomes the backtick form`he said "hi" it's`If a string somehow needs all three styles at once (it contains a double quote, a single quote, and a backtick), it is left unchanged, since no single style can hold it without escaping.
How
A
_normalize_quotes/_decode_literal/_encode_literaltrio insrc/cfengine_cli/format.pydecodes a literal to its logical content (honoring CFEngine's escaping rules — single/double quotes recognize only\\,\",\', and a backslash-newline line continuation; backticks process no escapes), chooses the preferred style from that content, and re-encodes. It is applied at thequoted_stringchoke point (stringify_single_line_node) and the promiser site (_promiser_text), covered by thetests/format/012_quotesgolden fixture.Note for review
Promisers are normalized too (e.g.
reports: 'hello';becomesreports: "hello";). Calling this out explicitly so it can be vetoed if promisers should be left alone.Generated by Claude Code