# Changelog
All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## Unreleased
## [0.8.1] 2026-07-26
### Added
- Added `Token::into_encoded`, which consumes the `Token` and returns its
encoded string representation as a `Cow<'a, str>` without allocating.
Resolves [#115](https://github.com/chanced/jsonptr/issues/115).
- `ReplaceError` (returned by `PointerBuf::replace`) is now re-exported from
the crate root, so callers can actually name the type.
### Changed
- Resolved all `clippy` (`clippy::all` + `clippy::pedantic`) and `rustc` lint
warnings across the full feature powerset — including newer-toolchain lints
(`mismatched_lifetime_syntaxes`) and feature-combination dead-code warnings.
Internal only; no public API changes.
### Fixed
- Fixed several broken/stale intra-doc links (a `crate::asign` typo, an "RFC
6091" typo, stale `toml` docs.rs links pointing at 0.8 instead of 0.9) and
a doc comment on `Pointer::len` that used `//` instead of `///` and was
silently dropped from rendered docs.
## [0.8.0] 2026-07-26
### Added
- Adds `Pointer::split_at_offset`, the byte-offset-based replacement for the
now-deprecated `Pointer::split_at`.
### Deprecated
- Deprecated `ResolveError` name.
- Deprecated `Pointer::split_at` in favor of `Pointer::split_at_offset`. The
`split_at` name is being reserved so it can be reintroduced as a position
(index) based split by 1.0, consistent with the rest of the API (e.g.
`Pointer::get`).
### Fixed
- `Delete::delete` no longer panics when a pointer's final token resolves to
an array index equal to the array's length (e.g. `-` on any array, or a
numeric index equal to the length). It now returns `None` and leaves the
document unmodified, as documented. Fixes
[#120](https://github.com/chanced/jsonptr/issues/120).
- `EncodingError` and `ParseIndexError` now implement `Diagnostic`, which
unifies the API for errors originating from parse-like operations.
- Fixes returning an incorrect error type when parsing a `Token` that
terminates with `~`. This now correctly classifies the error as a `Tilde`
error.
- Disabled default features for the `serde` dependency, which enables using
this library in no-std environments even with the `serde` feature enabled.
### Removed
- Some methods were removed from `InvalidCharacterError`, as that type no
longer holds a copy of the input string internally. This is a breaking
change. To access equivalent functionality, use the `Diagnostic` API
integration.
### Changed
- `Token::from_encoded` now accepts owned or borrowed strings (any type that
implements `Into<Cow<'_, str>>`).
- Sealed the `Diagnose` trait.
- Implementation of the `Default` trait for `Pointer` now doesn't constrain the lifetime.
## [0.7.1] 2025-02-16
### Changed
- Removes accidentally enabled default features `"miette"` and `"toml"`
## [0.7.0] 2025-02-13
### Added
- Adds method `into_buf` for `Box<Pointer>` and `impl From<PathBuf> for Box<Pointer>`.
- Adds unsafe associated methods `Pointer::new_unchecked` and `PointerBuf::new_unchecked` for
external zero-cost construction.
- Adds `Pointer::starts_with` and `Pointer::ends_with` for prefix and suffix matching.
- Adds new `ParseIndexError` variant to express the presence non-digit characters in the token.
- Adds `Token::is_next` for checking if a token represents the `-` character.
- Adds `InvalidEncoding` to represent the two possible encoding errors when decoding a token.
- Adds `diagnotic::Diagnostic` trait to facilitate error reporting and
`miette` integration. All errors intended for usage with `assign::Assign` or
`resolve::Resolve` must implement this trait.
- Adds `diagnostic::Report<T>` to capture the input for `PointerBuf::parse`
and to facilitate `miette` integration for all errors.
- Adds `"miette"` feature flag to enable `miette` integration for error reporting.
### Changed
- `Pointer::get` now accepts ranges and can produce `Pointer` segments as output (similar to
`slice::get`).
- Bumps minimum Rust version to 1.79.
- `PointerBuf::parse` now returns `RichParseError`, an alias to
`Report<ParseError>` which contains the allocated string as well as the
error. Use `Report::original` for matches or `Report::
- Renames `ParseError::NoLeadingBackslash` to `ParseError::NoLeadingSlash`
(sorry for the churn, I spaced hard - @chanced).
- Adds field `position` to variants of `resolve::Error` and `assign::Error` to indicate the
token index of where the error occurred.
- Renames `ParseError::is_no_leading_backslash` to `ParseError::is_no_leading_slash`.
- Renames `assign::AssignError` to `assign::Error`
- Renames `resolve::ResolveError` to `resolve::Error`
- Renames `InvalidEncodingError` to `EncodingError`
### Fixed
- Make validation of array indices conform to RFC 6901 in the presence of non-digit characters.
### Deprecated
- `ParseError::is_no_leading_backslash` renamed to `ParseError::is_no_leading_slash`.
- `assign::AssignError` renamed to `assign::Error`
- `resolve::ResolveError` renamed to `resolve::Error`
- `InvalidEncodingError` renamed to `EncodingError`
## [0.6.2] 2024-09-30
### Added
- Adds methods `len` and `is_empty` to `Pointer`
## [0.6.1] 2024-09-26
## Added
- Adds fluid methods `with_trailing_token`, `with_leading_token`, `concat` to `Pointer`.
## [0.6.0] - 2024-08-06
### Fixed
- `Token::to_index` now fails if the token contains leading zeros, as mandated by the RFC.
### Changed
- `ParseIndexError` is now an enum to reflect the new failure mode when parsing indices.
## [0.5.1]
### Changed
- README tweak.
## [0.5.0]
This is a breaking release including:
- [#30](https://github.com/chanced/jsonptr/pull/30) and [#37](https://github.com/chanced/jsonptr/pull/37) by [@asmello](https://github.com/asmello)
- [#41](https://github.com/chanced/jsonptr/pull/41) by [@chanced](https://github.com/chanced) & [@asmello](https://github.com/asmello)
### Added
- New slice type `Pointer` that enables zero-copy usage patterns
- New const constructor `const fn Pointer::from_static` for compile-time allocated `Pointer`s
- Zero-allocation `Pointer::root` singleton pointer
- [Quickcheck](https://docs.rs/quickcheck/latest/quickcheck/index.html)-based testing
- New methods: `Pointer::split_front`, `Pointer::split_back`, `Pointer::parent`, `Pointer::strip_suffix`
- Implemented `Display` and `Debug` for `ParseError`
- Adds `Pointer::split_at` which utilizes character offsets to split a pointer at a separator
- Adds specific error types `ParseError`, `ResolveError`, `AssignError`
### Changed
- JSON Pointers with leading `"#"` are no longer accepted. Previously, the erroneous leading hashtag was allowed during parsing but discarded.
- `Assign`, `Resolve`, `ResolveMut`, `Delete` all now use associated types `Value` and `Error`, allowing for more impls other than JSON
- Debug implementation now preserves type information (e.g. prints `PathBuf("/foo/bar")` instead of `"/foo/bar"`) - `Display` remains the same
- Original `Pointer` type renamed to `PointerBuf`
- Error types now use character `offset` indexing instead of owned copies of `Pointer` and `Token`.
- `Pointer::root` is now `PointerBuf::new`
- `Pointer::new` is now `PointerBuf::from_tokens` (and takes an `IntoIterator` argument - arrays still work)
- `Pointer::union` is now `PointerBuf::intersection`
- `Token` type has been simplified and is now by default a borrowed type (use `Token::to_owned` or `Token::into_owned` to make it owned)
- `Assign::assign` now returns `Result<Option<Assign::Value>, AssignError>`, where `Option<Assign::Value>` is the replaced value
### Fixed
- Fixes [#28](https://github.com/chanced/jsonptr/pull/28): `Pointer::union` is misleadingly named
### Removed
- Removes `Assignment`
- Removes `MaybePointer`
- Removes `Error`
- Removes `impl Deref<Target=&str>` from `Pointer`
- Removes optional dependencies of `url`, `fluent-uri` and `uniresid` as well as the `TryFrom` implementations for their respective types
- Removed `Token::as_key` and `Token::as_str` - use `Token::decoded().as_ref()` to achieve the same effect
- Several redundant or error-prone trait implementations were removed from `Token`
## [0.4.7] 2024-03-18
- Fixes issue with `pop_front` on a token with an empty string leaving the pointer in an invalid state. #25 by [@wngr](https://github.com/wngr)
- Fixes issue with `pop_back` on a token with an empty string. #26 by [@asmello](https://github.com/asmello)
## [0.4.6] 2024-03-24
- Fixes `Pointer::last` panicking for empty/root pointers #23 by [@wngr](https://github.com/wngr)
## [0.4.5] 2024-02-23
### Fixed
- Fixes issue with `Pointer::push_back` that does not allow for empty strings
to be appended as tokens. #21 fixed by [@wngr](https://github.com/wngr)
## [0.4.3] 2023-08-20
### Added
- Adds `parse` method to `Pointer` which calls the currently existing `FromStr`
impl
## [0.4.2] 2023-06-23
### Added
- implements `IntoIterator` for `&Pointer`
## [0.4.1] 2023-06-21
### Added
- implements `Borrow<[u8]>` and `AsRef<[u8]>` for `Pointer`
## [0.4.0] 2023-05-31
### Added
- Adds `CHANGELOG.md` which will be better upkept moving forward.
- Adds `MaybePointer` to assist with deserialization which should not fail fast.
### Changed
- `Pointer::new` now accepts a generic list, so `&["example"]` can be replaced by `["example"]`. For untyped, empty slices (i.e. `Pointer::new(&[])`), use `Pointer::default()`.
- `std` is now enabled by default.
### Removed
- Removes optional `MalformedPointerError` from `Pointer`.
## [0.3.6] 2023-05-23
### Changed
- Adds quotes around `Pointer` debug output (#11)
### Fixed
- Adds missing `impl std::error::Error` for `Error`, `NotFoundError`, `MalformedError`
- Fixes build for `std` feature flag
## [0.3.4] 2023-05-11
### Added
- Adds feature flag `fluent-uri` for `From<fluent_uri::Uri<_>` impl (#3)
## [0.2.0] 2023-02-24
### Changed
- `std` is now optional
- Adds feature flags `"uniresid"`, `"url"` to enable implementing `From<Uri>`, `From<Url>` (respectively).
### Removed
- Removes `Cargo.lock`
- Makes `uniresid` and `uri` optional
## [0.1.0] - 2022-06-12
### Fixed
- Fixes root pointer representation `""` rather than the erroneous `"/"`
- Fixes an issue where encoded tokens were not being resolved properly