GitoxideLabs/gitoxide gix-config-v0.59.0

gix-config v0.59.0

on GitHub

New Features

• add parse::format to normalize git-config whitespace.
  Adds gix_config::parse::format::normalize(input, &Options), which re-emits a
  single config file with sanitized whitespace without resolving includes. Values,
  comments and section headers are preserved verbatim; only insignificant
  whitespace, the = separator and newlines are rewritten.

  Options controls indentation (two spaces by default; tabs or none also
  available), spacing around =, the newline sequence (detect/LF/CRLF), the
  trailing newline, and optional blank-line collapsing.

  This is the library portion; the CLI will follow separately.

• return config values with their containing sections via File::*_with_sections() methods.
  That way it's easier to know where a value is coming from.
  Also streamline arguments and available methods for concistency.

• add SectionMut::rename() for precise single-section renames.
  This is important now that File::rename_section*() renames all matching
  sections.

Bug Fixes

• rename all matching sections in File::rename_section(), instead of just the last matching one.
  This is the correct behaviour and is what Git uses, even though it could break previous users in a setting
  where there are global configurations, like remote sections.

Changed (BREAKING)

• validate config value names internally and use AsBStrOpt for value pushes.
  This makes using the APIs much more convenient.

Chore (BREAKING)

• adapt to changes in gix-config
  Marked as breaking explicitly to as we know paramter types were
  changed to use convenience traits.

New Features (BREAKING)

• make parsed configuration files lifetime-free
  Store parsed configuration data in a self-contained backing buffer, with
  event payloads represented as compact spans into it. This removes the
  input lifetime from File and Events, making parsed configurations
  easier to return, store, move, and combine.

  Expose data through lightweight EventRef, SectionRef, HeaderRef,
  and BodyRef views. This retains efficient borrowed access while keeping
  ownership details out of the primary API.

  • Unify parsing under Events::from_bytes() and remove
    from_bytes_owned().
  • Add Events::iter(), frontmatter(), and sections() for
    non-consuming traversal.
  • Return owned BString values from lookups, avoiding Cow handling.
  • Simplify optional typed lookups from Option<Result<T, E>> to
    Result<Option<T>, E>.
  • Accept common string and byte-string inputs through AsBStr and
    IntoBStringOpt.
  • Allow concise calls such as new_section("remote", "origin") without
    constructing Option<Cow<...>>.
  • Reduce mutable value and section handles to the lifetime of their
    active file borrow.

  Introduce explicit section ownership: Section is self-contained and
  can be removed, edited, and inserted into another file, while
  SectionRef and SectionMut provide borrowed access.

  Preserve original subsection spelling for serialization while retaining
  decoded names for lookup. Configuration equality now also compares
  section and value names case-insensitively, matching Git semantics.

  Add benchmarks covering section, value, subsection, and multi-value
  lookups across configurations of varying sizes.

Bug Fixes (BREAKING)

• simplify gix-config section lookup acceleration
  breaking change for safety, as parameter types were changed.

  --- agent

  Replace the terminal/non-terminal enum vector with a direct boxed lookup
  record, use a Vec for section order, and centralize ordered insertion and
  removal. Empty lookup buckets are removed instead of becoming special cases.

  This removes 137 production lines, including 108 lines from the lookup
  utility, while adding targeted ordering and cleanup coverage.

  Back-to-back lookup benchmarks against the previous revision show:

  • section-name lookup: +0.4% at 10 and 100 sections, +4.2% at 1,000
  • subsection lookup: 5.7% to 9.1% faster
  • all-sections lookup: between 5.5% faster and 4.7% slower
  • isolated complete-value lookup: +0.8%, with no significant change

  Thus all lookup groups remain within the 5% parity target, with subsection
  lookup becoming consistently faster.

Commit Statistics

• 21 commits contributed to the release.
• 31 days passed between releases.
• 8 commits were understood as conventional.
• 1 unique issue was worked on: #2594

Thanks Clippy

Clippy helped 1 time to make code idiomatic.

Commit Details