diff --git a/Cargo.lock b/Cargo.lock index 47d8945ec56..5912a30127d 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -2053,9 +2053,11 @@ dependencies = [ "gix-object 0.44.0", "gix-path 0.10.11", "gix-tempfile 14.0.2", + "gix-testtools", "gix-trace 0.1.10", "gix-worktree 0.36.0", "imara-diff", + "pretty_assertions", "serde", "thiserror", ] diff --git a/crate-status.md b/crate-status.md index ad686714572..c0e24adabf2 100644 --- a/crate-status.md +++ b/crate-status.md @@ -324,19 +324,23 @@ Check out the [performance discussion][gix-diff-performance] as well. * [x] prepare invocation of external diff program - [ ] pass meta-info * [ ] working with hunks of data +* [ ] diff-heuristics match Git perfectly * [x] API documentation * [ ] Examples +[gix-diff-performance]: https://github.com/Byron/gitoxide/discussions/74 + ### gix-merge -* [ ] three-way merge analysis of blobs with choice of how to resolve conflicts +* [x] three-way merge analysis of blobs with choice of how to resolve conflicts - [ ] choose how to resolve conflicts on the data-structure - [ ] produce a new blob based on data-structure containing possible resolutions - - [ ] `merge` style - - [ ] `diff3` style - - [ ] `zdiff` style - -[gix-diff-performance]: https://github.com/Byron/gitoxide/discussions/74 + - [x] `merge` style + - [x] `diff3` style + - [x] `zdiff` style +* [ ] diff-heuristics match Git perfectly +* [x] API documentation + * [ ] Examples ### gix-traverse diff --git a/gix-merge/Cargo.toml b/gix-merge/Cargo.toml index b75d4cb3845..6d8da010147 100644 --- a/gix-merge/Cargo.toml +++ b/gix-merge/Cargo.toml @@ -39,6 +39,10 @@ serde = { version = "1.0.114", optional = true, default-features = false, featur document-features = { version = "0.2.0", optional = true } +[dev-dependencies] +gix-testtools = { path = "../tests/tools" } +pretty_assertions = "1.4.0" + [package.metadata.docs.rs] all-features = true features = ["document-features"] diff --git a/gix-merge/src/blob/builtin_driver.rs b/gix-merge/src/blob/builtin_driver.rs index cacef327ac6..b20fb04f096 100644 --- a/gix-merge/src/blob/builtin_driver.rs +++ b/gix-merge/src/blob/builtin_driver.rs @@ -23,8 +23,6 @@ impl BuiltinDriver { /// pub mod binary { - use crate::blob::Resolution; - /// What to do when having to pick a side to resolve a conflict. #[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] pub enum ResolveWith { @@ -36,7 +34,7 @@ pub mod binary { Theirs, } - /// Tell the caller of [`merge()`] which side was picked + /// Tell the caller of [`merge()`](function::merge) which side was picked. #[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] pub enum Pick { /// Chose the ancestor. @@ -47,23 +45,32 @@ pub mod binary { Theirs, } - /// As this algorithm doesn't look at the actual data, it returns a choice solely based on logic. - /// - /// It always results in a conflict with `current` being picked unless `on_conflict` is not `None`. - pub fn merge(on_conflict: Option) -> (Pick, Resolution) { - match on_conflict { - None => (Pick::Ours, Resolution::Conflict), - Some(ResolveWith::Ours) => (Pick::Ours, Resolution::Complete), - Some(ResolveWith::Theirs) => (Pick::Theirs, Resolution::Complete), - Some(ResolveWith::Ancestor) => (Pick::Ancestor, Resolution::Complete), + pub(super) mod function { + use crate::blob::builtin_driver::binary::{Pick, ResolveWith}; + use crate::blob::Resolution; + + /// As this algorithm doesn't look at the actual data, it returns a choice solely based on logic. + /// + /// It always results in a conflict with `current` being picked unless `on_conflict` is not `None`. + pub fn merge(on_conflict: Option) -> (Pick, Resolution) { + match on_conflict { + None => (Pick::Ours, Resolution::Conflict), + Some(resolve) => ( + match resolve { + ResolveWith::Ours => Pick::Ours, + ResolveWith::Theirs => Pick::Theirs, + ResolveWith::Ancestor => Pick::Ancestor, + }, + Resolution::Complete, + ), + } } } } +pub use binary::function::merge as binary; /// pub mod text { - use crate::blob::Resolution; - /// The way the built-in [text driver](crate::blob::BuiltinDriver::Text) will express /// merge conflicts in the resulting file. #[derive(Default, Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] @@ -115,8 +122,11 @@ pub mod text { } /// Options for the builtin [text driver](crate::blob::BuiltinDriver::Text). - #[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] + #[derive(Copy, Clone, Debug, Eq, PartialEq)] pub struct Options { + /// Determine of the diff will be performed. + /// Defaults to [`imara_diff::Algorithm::Myers`]. + pub diff_algorithm: imara_diff::Algorithm, /// How to visualize conflicts in merged files. pub conflict_style: ConflictStyle, /// The amount of markers to draw, defaults to 7, i.e. `<<<<<<<` @@ -132,6 +142,7 @@ pub mod text { conflict_style: Default::default(), marker_size: 7, on_conflict: None, + diff_algorithm: imara_diff::Algorithm::Myers, } } } @@ -147,10 +158,763 @@ pub mod text { Union, } - /// Merge `current` and `other` with `ancestor` as base according to `opts`. - /// - /// Place the merged result in `out` and return the resolution. - pub fn merge(_out: &mut Vec, _current: &[u8], _ancestor: &[u8], _other: &[u8], _opts: Options) -> Resolution { - todo!("text merge"); + pub(super) mod function { + use crate::blob::builtin_driver::text::{ConflictStyle, Options, ResolveWith}; + use crate::blob::Resolution; + use bstr::{BStr, ByteSlice, ByteVec}; + use std::iter::Peekable; + use std::ops::Range; + + /// Merge `current` and `other` with `ancestor` as base according to `opts`. + /// Use `current_label`, `other_label` and `ancestor_label` to annotate conflict sections. + /// + /// `input` is for reusing memory for lists of tokens, but note that it grows indefinitely + /// while tokens for `current`, `ancestor` and `other` are added. + /// Place the merged result in `out` (cleared before use) and return the resolution. + /// + /// # Important + /// + /// *The caller* is responsible for clearing it, otherwise tokens will accumulate. + /// This idea is to save time if the input is known to be very similar. + #[allow(clippy::too_many_arguments)] + pub fn merge<'a>( + out: &mut Vec, + input: &mut imara_diff::intern::InternedInput<&'a [u8]>, + current: &'a [u8], + current_label: Option<&BStr>, + ancestor: &'a [u8], + ancestor_label: Option<&BStr>, + other: &'a [u8], + other_label: Option<&BStr>, + opts: Options, + ) -> Resolution { + out.clear(); + input.update_before(tokens(ancestor)); + input.update_after(tokens(current)); + + let current_hunks = imara_diff::diff( + opts.diff_algorithm, + input, + CollectHunks { + side: Side::Current, + hunks: Default::default(), + }, + ); + + let current_tokens = std::mem::take(&mut input.after); + input.update_after(tokens(other)); + + let mut hunks = imara_diff::diff( + opts.diff_algorithm, + input, + CollectHunks { + side: Side::Other, + hunks: current_hunks, + }, + ); + + hunks.sort_by(|a, b| a.before.start.cmp(&b.before.start)); + let mut hunks = hunks.into_iter().peekable(); + let mut intersecting = Vec::new(); + let mut ancestor_integrated_until = 0; + let mut resolution = Resolution::Complete; + // TODO(performance): instead of skipping hunks, let's not compute these ones at all, but only once all tests are there. + let mut filled_hunks = Vec::with_capacity(2); + while let Some(hunk) = hunks.next() { + if take_intersecting(&hunk, &mut hunks, &mut intersecting) { + fill_ancestor(&hunk.before, &mut intersecting); + + let filled_hunks_side = hunk.side; + filled_hunks.clear(); + filled_hunks.push(hunk); + fill_ancestor( + &intersecting + .first() + .zip(intersecting.last()) + .map(|(f, l)| f.before.start..l.before.end) + .expect("at least one entry"), + &mut filled_hunks, + ); + match opts.on_conflict { + None => { + let (hunks_front_and_back, num_hunks_front) = match opts.conflict_style { + ConflictStyle::Merge | ConflictStyle::ZealousDiff3 => zealously_contract_hunks( + &mut filled_hunks, + &mut intersecting, + input, + ¤t_tokens, + ), + ConflictStyle::Diff3 => (Vec::new(), 0), + }; + let (our_hunks, their_hunks) = match filled_hunks_side { + Side::Current => (&filled_hunks, &intersecting), + Side::Other => (&intersecting, &filled_hunks), + Side::Ancestor => { + unreachable!("initial hunks are never ancestors") + } + }; + // TODO: dedup preamble, simplify this - we know that our and their hunks aren't empty. + let (front_hunks, back_hunks) = hunks_front_and_back.split_at(num_hunks_front); + let first_hunk = front_hunks + .first() + .or(our_hunks.first()) + .expect("at least one hunk to write"); + let last_hunk = back_hunks + .last() + .or(their_hunks.last()) + .or(our_hunks.last()) + .or(front_hunks.last()) + .expect("at least one hunk"); + write_ancestor(input, ancestor_integrated_until, first_hunk.before.start as usize, out); + + write_hunks(front_hunks, input, ¤t_tokens, out); + if their_hunks.is_empty() { + write_hunks(our_hunks, input, ¤t_tokens, out); + } else if our_hunks.is_empty() { + // TODO: assure we run into this - currently no test triggers this. + write_hunks(their_hunks, input, ¤t_tokens, out); + } else { + // DEVIATION: this makes tests (mostly) pass, but probably is very different from what Git does. + let hunk_storage; + let nl = detect_line_ending( + if front_hunks.is_empty() { + hunk_storage = Hunk { + before: ancestor_integrated_until..first_hunk.before.start, + after: Default::default(), + side: Side::Ancestor, + }; + std::slice::from_ref(&hunk_storage) + } else { + front_hunks + }, + input, + ¤t_tokens, + ) + .or_else(|| detect_line_ending(our_hunks, input, ¤t_tokens)) + .unwrap_or(b"\n".into()); + match opts.conflict_style { + ConflictStyle::Merge => { + if contains_lines(our_hunks) || contains_lines(their_hunks) { + resolution = Resolution::Conflict; + write_conflict_marker(out, b'<', current_label, opts.marker_size, nl); + write_hunks(our_hunks, input, ¤t_tokens, out); + write_conflict_marker(out, b'=', None, opts.marker_size, nl); + write_hunks(their_hunks, input, ¤t_tokens, out); + write_conflict_marker(out, b'>', other_label, opts.marker_size, nl); + } + } + ConflictStyle::Diff3 | ConflictStyle::ZealousDiff3 => { + if contains_lines(our_hunks) || contains_lines(their_hunks) { + if hunks_differ_in_diff3( + opts.conflict_style, + our_hunks, + their_hunks, + input, + ¤t_tokens, + ) { + resolution = Resolution::Conflict; + write_conflict_marker(out, b'<', current_label, opts.marker_size, nl); + write_hunks(our_hunks, input, ¤t_tokens, out); + let ancestor_hunk = Hunk { + before: first_hunk.before.start..last_hunk.before.end, + after: Default::default(), + side: Side::Ancestor, + }; + let ancestor_hunk = std::slice::from_ref(&ancestor_hunk); + let ancestor_nl = + detect_line_ending_or_nl(ancestor_hunk, input, ¤t_tokens); + write_conflict_marker( + out, + b'|', + ancestor_label, + opts.marker_size, + ancestor_nl, + ); + write_hunks(ancestor_hunk, input, ¤t_tokens, out); + write_conflict_marker(out, b'=', None, opts.marker_size, nl); + write_hunks(their_hunks, input, ¤t_tokens, out); + write_conflict_marker(out, b'>', other_label, opts.marker_size, nl); + } else { + write_hunks(our_hunks, input, ¤t_tokens, out); + } + } + } + } + } + write_hunks(back_hunks, input, ¤t_tokens, out); + ancestor_integrated_until = last_hunk.before.end; + } + Some(resolve) => { + match resolve { + ResolveWith::Ours | ResolveWith::Theirs => { + let (our_hunks, their_hunks) = match filled_hunks_side { + Side::Current => (&filled_hunks, &intersecting), + Side::Other => (&intersecting, &filled_hunks), + Side::Ancestor => { + unreachable!("initial hunks are never ancestors") + } + }; + let hunks_to_write = if resolve == ResolveWith::Ours { + our_hunks + } else { + their_hunks + }; + if let Some(first_hunk) = hunks_to_write.first() { + write_ancestor( + input, + ancestor_integrated_until, + first_hunk.before.start as usize, + out, + ); + } + write_hunks(hunks_to_write, input, ¤t_tokens, out); + if let Some(last_hunk) = hunks_to_write.last() { + ancestor_integrated_until = last_hunk.before.end; + } + } + ResolveWith::Union => { + let (hunks_front_and_back, num_hunks_front) = zealously_contract_hunks( + &mut filled_hunks, + &mut intersecting, + input, + ¤t_tokens, + ); + + let (our_hunks, their_hunks) = match filled_hunks_side { + Side::Current => (&filled_hunks, &intersecting), + Side::Other => (&intersecting, &filled_hunks), + Side::Ancestor => { + unreachable!("initial hunks are never ancestors") + } + }; + let (front_hunks, back_hunks) = hunks_front_and_back.split_at(num_hunks_front); + let first_hunk = front_hunks + .first() + .or(our_hunks.first()) + .expect("at least one hunk to write"); + write_ancestor( + input, + ancestor_integrated_until, + first_hunk.before.start as usize, + out, + ); + write_hunks(front_hunks, input, ¤t_tokens, out); + assure_ends_with_nl( + out, + detect_line_ending_or_nl(front_hunks, input, ¤t_tokens), + ); + write_hunks(our_hunks, input, ¤t_tokens, out); + assure_ends_with_nl( + out, + detect_line_ending_or_nl(our_hunks, input, ¤t_tokens), + ); + write_hunks(their_hunks, input, ¤t_tokens, out); + if !back_hunks.is_empty() { + assure_ends_with_nl( + out, + detect_line_ending_or_nl(their_hunks, input, ¤t_tokens), + ); + } + write_hunks(back_hunks, input, ¤t_tokens, out); + let last_hunk = back_hunks + .last() + .or(their_hunks.last()) + .or(our_hunks.last()) + .or(front_hunks.last()) + .expect("at least one hunk"); + ancestor_integrated_until = last_hunk.before.end; + } + }; + } + } + } else { + write_ancestor(input, ancestor_integrated_until, hunk.before.start as usize, out); + ancestor_integrated_until = hunk.before.end; + write_hunks(std::slice::from_ref(&hunk), input, ¤t_tokens, out); + } + } + write_ancestor(input, ancestor_integrated_until, input.before.len(), out); + + resolution + } + + /// Used only when `diff3` is the conflict style as `zdiff3` automatically reduces hunks into nothing. + /// Here we check if all hunks are the same. + fn hunks_differ_in_diff3( + style: ConflictStyle, + a: &[Hunk], + b: &[Hunk], + input: &imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &[imara_diff::intern::Token], + ) -> bool { + if style != ConflictStyle::Diff3 { + return true; + } + + let tokens_for_hunk = + |hunk: &Hunk| -> &[imara_diff::intern::Token] { tokens_for_side(hunk.side, input, current_tokens) }; + + a.iter() + .flat_map(tokens_for_hunk) + .ne(b.iter().flat_map(tokens_for_hunk)) + } + + fn contains_lines(hunks: &[Hunk]) -> bool { + hunks.iter().any(|h| !h.after.is_empty()) + } + + /// ## Deviation + /// + /// This implementation definitely isn't the same as in Git, primarily because it seemed impossible + /// to understand what's going on there without investing more time than it seemed worth. + fn detect_line_ending( + hunks: &[Hunk], + input: &mut imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &[imara_diff::intern::Token], + ) -> Option<&'static BStr> { + fn is_eol_crlf( + hunks: &[Hunk], + input: &mut imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &[imara_diff::intern::Token], + ) -> Option { + let (range, side) = hunks.iter().rev().find_map(|h| { + (!h.after.is_empty()) + .then_some((&h.after, h.side)) + .or((!h.before.is_empty()).then_some((&h.before, Side::Ancestor))) + })?; + + let tokens = tokens_for_side(side, input, current_tokens); + { + let last_line = tokens + .get(range.end as usize - 1) + .map(|token| &input.interner[*token])?; + if last_line.last() == Some(&b'\n') { + return last_line.get(last_line.len().checked_sub(2)?).map(|c| *c == b'\r'); + } + } + let second_to_last_line = tokens + .get(range.end.checked_sub(2)? as usize) + .map(|token| &input.interner[*token])?; + second_to_last_line + .get(second_to_last_line.len().checked_sub(2)?) + .map(|c| *c == b'\r') + } + is_eol_crlf(hunks, input, current_tokens).map(|is_crlf| if is_crlf { b"\r\n".into() } else { b"\n".into() }) + } + + fn detect_line_ending_or_nl( + hunks: &[Hunk], + input: &mut imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &[imara_diff::intern::Token], + ) -> &'static BStr { + detect_line_ending(hunks, input, current_tokens).unwrap_or(b"\n".into()) + } + + fn tokens_for_side<'a>( + side: Side, + input: &'a imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &'a [imara_diff::intern::Token], + ) -> &'a [imara_diff::intern::Token] { + match side { + Side::Current => current_tokens, + Side::Other => &input.after, + Side::Ancestor => &input.before, + } + } + + fn assure_ends_with_nl(out: &mut Vec, nl: &BStr) { + if !out.is_empty() && !out.ends_with(b"\n") { + out.push_str(nl); + } + } + + fn write_conflict_marker(out: &mut Vec, marker: u8, label: Option<&BStr>, marker_size: usize, nl: &BStr) { + assure_ends_with_nl(out, nl); + out.extend(std::iter::repeat(marker).take(marker_size)); + if let Some(label) = label { + out.push(b' '); + out.extend_from_slice(label); + } + out.push_str(nl); + } + + fn write_ancestor(input: &imara_diff::intern::InternedInput<&[u8]>, from: u32, to: usize, out: &mut Vec) { + if to < from as usize { + return; + } + if let Some(tokens) = input.before.get(from as usize..to) { + write_tokens(&input.interner, tokens, out); + } + } + + /// Look at all hunks in `in_out` and fill in the ancestor in the range of `ancestor_range`. + /// This is all based on knowing the ranges are sequences of tokens. + fn fill_ancestor(Range { start, end }: &Range, in_out: &mut Vec) { + if in_out.is_empty() { + return; + } + + fn ancestor_hunk(start: u32, num_lines: u32) -> Hunk { + let range = start..start + num_lines; + Hunk { + before: range.clone(), + after: range, + side: Side::Ancestor, + } + } + + fn is_nonzero(num: &u32) -> bool { + *num > 0 + } + + let first = &in_out[0]; + let mut first_idx = 0; + if let Some(lines_to_add) = first.before.start.checked_sub(*start).filter(is_nonzero) { + in_out.insert(0, ancestor_hunk(*start, lines_to_add)); + first_idx += 1; + } + + let mut added_hunks = false; + for (idx, next_idx) in (first_idx..in_out.len()).map(|idx| (idx, idx + 1)) { + let Some(next_hunk) = in_out.get(next_idx) else { break }; + let hunk = &in_out[idx]; + if let Some(lines_to_add) = next_hunk.after.start.checked_sub(hunk.after.end).filter(is_nonzero) { + in_out.push(ancestor_hunk(hunk.after.end, lines_to_add)); + added_hunks = true; + } + } + let in_out_len = in_out.len(); + if added_hunks { + in_out[first_idx..in_out_len].sort_by_key(|hunk| hunk.before.start); + } + + let last = &in_out[in_out_len - 1]; + if let Some(lines_to_add) = end.checked_sub(last.before.end).filter(is_nonzero) { + in_out.push(ancestor_hunk(last.before.end, lines_to_add)); + } + } + + /// Reduce the area of `a_hunks` and the hunks in `b_hunks` so that only those lines that are + /// actually different remain. Note that we have to compare the resolved values, not only the tokens, + /// so `current_tokens` is expected to be known to the `input` (and its `interner`). + /// Hunks from all input arrays maybe removed in the process from the front and back, in case they + /// are entirely equal to what's in `hunk`. Note also that `a_hunks` and `b_hunks` are treated to be consecutive, + /// so [`fill_ancestor()`] must have been called beforehand, and are assumed to covert the same space in the + /// ancestor buffer. + /// Use `mode` to determine how hunks may be handled. + /// + /// Return a new vector of all the hunks that were removed from front and back, with partial hunks inserted, + /// along with the amount of hunks that go front, with the remaining going towards the back. + // TODO: refactor so hunks and their associated data can go into an array for easier handling. + #[must_use] + fn zealously_contract_hunks( + a_hunks: &mut Vec, + b_hunks: &mut Vec, + input: &imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &[imara_diff::intern::Token], + ) -> (Vec, usize) { + let line_content = |token_idx: u32, side: Side| { + let tokens = match side { + Side::Current => current_tokens, + Side::Other => &input.after, + Side::Ancestor => &input.before, + }; + &input.interner[tokens[token_idx as usize]] + }; + fn range_by_side(hunk: &mut Hunk) -> &mut Range { + match hunk.side { + Side::Current | Side::Other => &mut hunk.after, + Side::Ancestor => &mut hunk.before, + } + } + fn truncate_hunks_from_from_front( + hunks: &mut Vec, + hunks_to_remove_until_idx: Option, + hunk_token_equal_till: Option, + mut out_hunks: Option<&mut Vec>, + ) { + let Some(hunks_to_remove_until_idx) = hunks_to_remove_until_idx else { + assert!(hunk_token_equal_till.is_none()); + return; + }; + let mut last_index_to_remove = Some(hunks_to_remove_until_idx); + let hunk = &mut hunks[hunks_to_remove_until_idx]; + let range = range_by_side(hunk); + if let Some(hunk_token_equal_till) = hunk_token_equal_till { + let orig_start = range.start; + let new_start = hunk_token_equal_till + 1; + range.start = new_start; + if Range::::is_empty(range) { + range.start = orig_start; + } else if let Some(out) = out_hunks.as_deref_mut() { + last_index_to_remove = hunks_to_remove_until_idx.checked_sub(1); + let mut removed_hunk = hunk.clone(); + let new_range = range_by_side(&mut removed_hunk); + + new_range.start = orig_start; + new_range.end = new_start; + + out.push(removed_hunk); + } else { + last_index_to_remove = hunks_to_remove_until_idx.checked_sub(1); + } + } + if let Some(last_index_to_remove) = last_index_to_remove { + let mut current_idx = 0; + hunks.retain(|hunk| { + if current_idx > last_index_to_remove { + true + } else { + current_idx += 1; + if let Some(out) = out_hunks.as_deref_mut() { + out.push(hunk.clone()); + } + false + } + }); + } + } + + fn truncate_hunks_from_from_back( + hunks: &mut Vec, + remove_trailing_hunks_from_idx: Option, + hunk_token_equal_from: Option, + mut out_hunks: Option<&mut Vec>, + ) { + let Some(mut remove_trailing_hunks_from_idx) = remove_trailing_hunks_from_idx else { + assert!(hunk_token_equal_from.is_none()); + return; + }; + + let hunk = &mut hunks[remove_trailing_hunks_from_idx]; + let range = range_by_side(hunk); + if let Some(hunk_token_equal_from) = hunk_token_equal_from { + let orig_end = range.end; + let new_end = hunk_token_equal_from; + range.end = new_end; + if Range::::is_empty(range) { + range.end = orig_end; + } else if let Some(out) = out_hunks.as_deref_mut() { + remove_trailing_hunks_from_idx += 1; + let mut removed_hunk = hunk.clone(); + let new_range = range_by_side(&mut removed_hunk); + + new_range.start = new_end; + new_range.end = orig_end; + + out.push(removed_hunk); + } else { + remove_trailing_hunks_from_idx += 1; + } + } + if let Some(out) = out_hunks { + out.extend_from_slice(&hunks[remove_trailing_hunks_from_idx..]); + } + hunks.truncate(remove_trailing_hunks_from_idx); + } + + let (mut last_a_hunk_idx, mut last_b_hunk_idx) = (0, 0); + let (mut out, hunks_in_front) = { + let (mut remove_leading_a_hunks_from, mut remove_leading_b_hunks_from) = (None, None); + let (mut a_hunk_token_equal_till, mut b_hunk_token_equal_till) = (None, None); + for ((a_token_idx, a_hunk_idx, a_hunk_side), (b_token_idx, b_hunk_idx, b_hunk_side)) in + iterate_hunks(a_hunks).zip(iterate_hunks(b_hunks)) + { + let a_line = line_content(a_token_idx, a_hunk_side).as_bstr(); + let b_line = line_content(b_token_idx, b_hunk_side).as_bstr(); + + if last_a_hunk_idx != a_hunk_idx { + a_hunk_token_equal_till = None; + last_a_hunk_idx = a_hunk_idx; + } + if last_b_hunk_idx != b_hunk_idx { + b_hunk_token_equal_till = None; + last_b_hunk_idx = b_hunk_idx; + } + if a_line == b_line { + (remove_leading_a_hunks_from, remove_leading_b_hunks_from) = + (Some(a_hunk_idx), Some(b_hunk_idx)); + (a_hunk_token_equal_till, b_hunk_token_equal_till) = (Some(a_token_idx), Some(b_token_idx)); + } else { + break; + } + } + + let mut out = Vec::with_capacity(remove_leading_a_hunks_from.unwrap_or_else(|| { + if a_hunk_token_equal_till.is_some() { + 1 + } else { + 0 + } + })); + truncate_hunks_from_from_front( + a_hunks, + remove_leading_a_hunks_from, + a_hunk_token_equal_till, + Some(&mut out), + ); + truncate_hunks_from_from_front(b_hunks, remove_leading_b_hunks_from, b_hunk_token_equal_till, None); + let hunks_in_front = out.len(); + (out, hunks_in_front) + }; + + (last_a_hunk_idx, last_b_hunk_idx) = (0, 0); + { + let (mut remove_trailing_a_hunks_from, mut remove_trailing_b_hunks_from) = (None, None); + let (mut a_hunk_token_equal_from, mut b_hunk_token_equal_from) = (None, None); + for ((a_token_idx, a_hunk_idx, a_hunk_side), (b_token_idx, b_hunk_idx, b_hunk_side)) in + iterate_hunks_rev(a_hunks).zip(iterate_hunks_rev(b_hunks)) + { + let a_line = line_content(a_token_idx, a_hunk_side).as_bstr(); + let b_line = line_content(b_token_idx, b_hunk_side).as_bstr(); + + if last_a_hunk_idx != a_hunk_idx { + a_hunk_token_equal_from = None; + last_a_hunk_idx = a_hunk_idx; + } + if last_b_hunk_idx != b_hunk_idx { + b_hunk_token_equal_from = None; + last_b_hunk_idx = b_hunk_idx; + } + + if a_line == b_line { + (remove_trailing_a_hunks_from, remove_trailing_b_hunks_from) = + (Some(a_hunk_idx), Some(b_hunk_idx)); + (a_hunk_token_equal_from, b_hunk_token_equal_from) = (Some(a_token_idx), Some(b_token_idx)); + } else { + break; + } + } + + truncate_hunks_from_from_back( + a_hunks, + remove_trailing_a_hunks_from, + a_hunk_token_equal_from, + Some(&mut out), + ); + truncate_hunks_from_from_back(b_hunks, remove_trailing_b_hunks_from, b_hunk_token_equal_from, None); + } + + (out, hunks_in_front) + } + + /// Return an iterator over `(token_idx, hunk_idx, hunk_side)` from `hunks`. + fn iterate_hunks(hunks: &[Hunk]) -> impl Iterator + '_ { + hunks.iter().enumerate().flat_map(|(hunk_idx, hunk)| { + match hunk.side { + Side::Current | Side::Other => &hunk.after, + Side::Ancestor => &hunk.before, + } + .clone() + .map(move |idx| (idx, hunk_idx, hunk.side)) + }) + } + + /// Return a reverse iterator over `(token_idx, hunk_idx, hunk_side)` from `hunks`. + fn iterate_hunks_rev(hunks: &[Hunk]) -> impl Iterator + '_ { + hunks.iter().enumerate().rev().flat_map(|(hunk_idx, hunk)| { + match hunk.side { + Side::Current | Side::Other => &hunk.after, + Side::Ancestor => &hunk.before, + } + .clone() + .rev() + .map(move |idx| (idx, hunk_idx, hunk.side)) + }) + } + + fn write_hunks( + hunks: &[Hunk], + input: &imara_diff::intern::InternedInput<&[u8]>, + current_tokens: &[imara_diff::intern::Token], + out: &mut Vec, + ) { + for hunk in hunks { + let (tokens, range) = match hunk.side { + Side::Current => (current_tokens, &hunk.after), + Side::Other => (input.after.as_slice(), &hunk.after), + Side::Ancestor => (input.before.as_slice(), &hunk.before), + }; + write_tokens(&input.interner, &tokens[usize_range(range)], out); + } + } + + fn usize_range(range: &Range) -> Range { + range.start as usize..range.end as usize + } + + fn write_tokens( + interner: &imara_diff::intern::Interner<&[u8]>, + tokens: &[imara_diff::intern::Token], + out: &mut Vec, + ) { + for token in tokens { + out.extend_from_slice(interner[*token]); + } + } + + /// Find all hunks in `iter` which aren't from the same side as `hunk` and intersect with it. + /// Return `true` if `out` is non-empty after the operation, indicating overlapping hunks were found. + fn take_intersecting( + hunk: &Hunk, + iter: &mut Peekable>, + out: &mut Vec, + ) -> bool { + out.clear(); + while iter + .peek() + .filter(|b_hunk| { + b_hunk.side != hunk.side + && (hunk.before.contains(&b_hunk.before.start) + || (hunk.before.is_empty() && hunk.before.start == b_hunk.before.start)) + }) + .is_some() + { + out.extend(iter.next()); + } + !out.is_empty() + } + + fn tokens(input: &[u8]) -> imara_diff::sources::ByteLines<'_, true> { + imara_diff::sources::byte_lines_with_terminator(input) + } + + #[derive(Debug, Copy, Clone, Eq, PartialEq)] + enum Side { + Current, + Other, + /// A special marker that is just used to be able to mix-in hunks that only point to the ancestor. + /// Only `before` matters then. + Ancestor, + } + + #[derive(Debug, Clone)] + struct Hunk { + before: Range, + after: Range, + side: Side, + } + + struct CollectHunks { + hunks: Vec, + side: Side, + } + + impl imara_diff::Sink for CollectHunks { + type Out = Vec; + + fn process_change(&mut self, before: Range, after: Range) { + self.hunks.push(Hunk { + before, + after, + side: self.side, + }); + } + + fn finish(self) -> Self::Out { + self.hunks + } + } } } +pub use text::function::merge as text; diff --git a/gix-merge/src/blob/platform.rs b/gix-merge/src/blob/platform.rs index 497b9bf887e..6b6175ee408 100644 --- a/gix-merge/src/blob/platform.rs +++ b/gix-merge/src/blob/platform.rs @@ -137,7 +137,7 @@ pub mod merge { pub other: ResourceRef<'parent>, } - #[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] + #[derive(Copy, Clone, Debug, Eq, PartialEq)] pub struct Options { /// If `true`, the resources being merged are contained in a virtual ancestor, /// which is the case when merge bases are merged into one. diff --git a/gix-merge/tests/fixtures/generated-archives/text-baseline.tar b/gix-merge/tests/fixtures/generated-archives/text-baseline.tar new file mode 100644 index 00000000000..b7ffc607c76 Binary files /dev/null and b/gix-merge/tests/fixtures/generated-archives/text-baseline.tar differ diff --git a/gix-merge/tests/fixtures/text-baseline.sh b/gix-merge/tests/fixtures/text-baseline.sh new file mode 100644 index 00000000000..47e160c56af --- /dev/null +++ b/gix-merge/tests/fixtures/text-baseline.sh @@ -0,0 +1,659 @@ +#!/usr/bin/env bash +set -eu -o pipefail + +git init +rm -Rf .git/hooks + +function baseline() { + local ours=$DIR/${1:?1: our file}.blob; + local base=$DIR/${2:?2: base file}.blob; + local theirs=$DIR/${3:?3: their file}.blob; + local output=$DIR/${4:?4: the name of the output file}.merged; + + shift 4 + git merge-file --stdout "$@" "$ours" "$base" "$theirs" > "$output" || true + + echo "$ours" "$base" "$theirs" "$output" "$@" >> baseline.cases +} + +mkdir simple +(cd simple + echo -e "line1-changed-by-both\nline2-to-be-changed-in-incoming" > ours.blob + echo -e "line1-to-be-changed-by-both\nline2-to-be-changed-in-incoming" > base.blob + echo -e "line1-changed-by-both\nline2-changed" > theirs.blob +) + +# one big change includes multiple smaller ones +mkdir multi-change +(cd multi-change + cat < base.blob +0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +EOF + + cat < ours.blob +0 +1 +X +X +4 +5 +Y +Y +8 +Z +EOF + + cat < theirs.blob +T +T +T +T +T +T +T +T +T +T +EOF +) + +# a change with deletion/clearing our file +mkdir clear-ours +(cd clear-ours + cat < base.blob +0 +1 +2 +3 +4 +5 +EOF + + touch ours.blob + + cat < theirs.blob +T +T +T +T +T +EOF +) + +# a change with deletion/clearing their file +mkdir clear-theirs +(cd clear-theirs + cat < base.blob +0 +1 +2 +3 +4 +5 +EOF + + cat < ours.blob +O +O +O +O +O +EOF + + touch theirs.blob +) + +# differently sized changes +mkdir ours-2-lines-theirs-1-line +(cd ours-2-lines-theirs-1-line + cat < base.blob +0 +1 +2 +3 +4 +5 +EOF + + cat < ours.blob +0 +1 +X +X +4 +5 +EOF + + cat < theirs.blob +0 +1 +Y +3 +4 +5 +EOF +) + +# partial match +mkdir partial-match +(cd partial-match + cat < base.blob +0 +1 +2 +3 +4 +5 +EOF + + cat < ours.blob +0 +X1 +X2 +X3 +X4 +5 +EOF + + cat < theirs.blob +0 +X1 +2 +X3 +X4 +5 +EOF +) + +# based on 'unique merge base' from 'diff3-conflict-markers' +mkdir unique-merge-base-with-insertion +(cd unique-merge-base-with-insertion + cat < base.blob +1 +2 +3 +4 +5 +EOF + + # no trailing newline + echo -n $'1\n2\n3\n4\n5\n7' > ours.blob + echo -n $'1\n2\n3\n4\n5\nsix' > theirs.blob +) + +mkdir zdiff3-basic +(cd zdiff3-basic + cat < base.blob +1 +2 +3 +4 +5 +6 +7 +8 +9 +EOF + + cat < ours.blob +1 +2 +3 +4 +A +B +C +D +E +7 +8 +9 +EOF + + cat < theirs.blob +1 +2 +3 +4 +A +X +C +Y +E +7 +8 +9 +EOF +) + +mkdir zdiff3-middlecommon +(cd zdiff3-middlecommon + cat < base.blob +1 +2 +3 +AA +4 +5 +BB +6 +7 +8 +EOF + + cat < ours.blob +1 +2 +3 +CC +4 +5 +DD +6 +7 +8 +EOF + + cat < theirs.blob +1 +2 +3 +EE +4 +5 +FF +6 +7 +8 +EOF +) + + +mkdir zdiff3-interesting +(cd zdiff3-interesting + cat < base.blob +1 +2 +3 +4 +5 +6 +7 +8 +9 +EOF + + cat < ours.blob +1 +2 +3 +4 +A +B +C +D +E +F +G +H +I +J +7 +8 +9 +EOF + + cat < theirs.blob +1 +2 +3 +4 +A +B +C +5 +6 +G +H +I +J +7 +8 +9 +EOF +) + +mkdir zdiff3-evil +(cd zdiff3-evil + cat < base.blob +1 +2 +3 +4 +5 +6 +7 +8 +9 +EOF + + cat < ours.blob +1 +2 +3 +4 +X +A +B +C +7 +8 +9 +EOF + + cat < theirs.blob +1 +2 +3 +4 +Y +A +B +C +B +C +7 +8 +9 +EOF +) + +mkdir no-change-add +(cd no-change-add + + echo -e " line 1\n\n line 2" > base.blob + echo -e " line 1\n\n line in between\n\n line 2\n\n line in between\n\n line 3" > ours.blob + cp ours.blob theirs.blob +) + +mkdir no-change-remove +(cd no-change-remove + + echo -e " line 1\n\n line in between\n\n line 2\n\n line in between\n\n line 3" > base.blob + echo -e " line 1\n\n line 2" > ours.blob + cp ours.blob theirs.blob +) + +mkdir complex +(cd complex + cat <base.blob +Dominus regit me, +et nihil mihi deerit. +In loco pascuae ibi me collocavit, +super aquam refectionis educavit me; +animam meam convertit, +deduxit me super semitas jusitiae, +propter nomen suum. +EOF + + cat <new1.blob +Dominus regit me, +et nihil mihi deerit. +In loco pascuae ibi me collocavit, +super aquam refectionis educavit me; +animam meam convertit, +deduxit me super semitas jusitiae, +propter nomen suum. +Nam et si ambulavero in medio umbrae mortis, +non timebo mala, quoniam tu mecum es: +virga tua et baculus tuus ipsa me consolata sunt. +EOF + + cat <new2.blob +Dominus regit me, et nihil mihi deerit. +In loco pascuae ibi me collocavit, +super aquam refectionis educavit me; +animam meam convertit, +deduxit me super semitas jusitiae, +propter nomen suum. +EOF + + cat <new3.blob +DOMINUS regit me, +et nihil mihi deerit. +In loco pascuae ibi me collocavit, +super aquam refectionis educavit me; +animam meam convertit, +deduxit me super semitas jusitiae, +propter nomen suum. +EOF + + cat <new4.blob +Dominus regit me, et nihil mihi deerit. +In loco pascuae ibi me collocavit, +super aquam refectionis educavit me; +animam meam convertit, +deduxit me super semitas jusitiae, +EOF + + cat <new5.blob +Dominus regit me, +et nihil mihi deerit. +In loco pascuae ibi me collocavit, +super aquam refectionis educavit me; +animam meam convertit, +deduxit me super semitas jusitiae, +propter nomen suum. +Nam et si ambulavero in medio umbrae mortis, +non timebo mala, quoniam TU mecum es: +virga tua et baculus tuus ipsa me consolata sunt. +EOF + + echo -n "propter nomen suum." >>new4.blob + + cat <base.c +int f(int x, int y) +{ + if (x == 0) + { + return y; + } + return x; +} + +int g(size_t u) +{ + while (u < 30) + { + u++; + } + return u; +} +EOF + + cat <ours.c +int g(size_t u) +{ + while (u < 30) + { + u++; + } + return u; +} + +int h(int x, int y, int z) +{ + if (z == 0) + { + return x; + } + return y; +} +EOF + + cat <theirs.c +int f(int x, int y) +{ + if (x == 0) + { + return y; + } + return x; +} + +int g(size_t u) +{ + while (u > 34) + { + u--; + } + return u; +} +EOF + + mkdir no-change + (cd no-change + cp ../base.blob . + cp base.blob ours.blob + cp base.blob theirs.blob + ) + mkdir no-conflict + (cd no-conflict + cp ../new1.blob ours.blob + cp ../base.blob base.blob + cp ../new2.blob theirs.blob + ) + mkdir no-conflict-too + (cd no-conflict-too + cp ../base.blob ours.blob + cp ../base.blob base.blob + cp ../new2.blob theirs.blob + ) + mkdir they-changed + (cd they-changed + touch ours.blob base.blob + cp ../new2.blob theirs.blob + ) + mkdir missing-LF-at-EOF + (cd missing-LF-at-EOF + cp ../new1.blob ours.blob + cp ../base.blob base.blob + cp ../new4.blob theirs.blob + ) + mkdir missing-LF-at-EOF-no-conflict + (cd missing-LF-at-EOF-no-conflict + cp ../new4.blob ours.blob + cp ../new2.blob base.blob + cp ../new3.blob theirs.blob + ) + mkdir with-conflicts + (cd with-conflicts + cp ../new1.blob ours.blob + cp ../base.blob base.blob + cp ../new3.blob theirs.blob + ) + mkdir with-conflicts-in-removed-tail + (cd with-conflicts-in-removed-tail + cp ../base.blob ours.blob + cp ../new1.blob base.blob + cp ../new5.blob theirs.blob + ) + mkdir auto-simplification + (cd auto-simplification + sed -e "s/deerit.\$/deerit;/" -e "s/me;\$/me./" <../new5.blob >ours.blob + cp ../new5.blob base.blob + sed -e "s/deerit.\$/deerit,/" -e "s/me;\$/me,/" <../new5.blob >theirs.blob + ) + mkdir auto-simplification2 + (cd auto-simplification2 + sed -e "s/deerit./&%%%%/" -e "s/locavit,/locavit;/" <../auto-simplification/ours.blob | tr % "\012" >ours.blob + cp ../new5.blob base.blob + sed -e "s/deerit./&%%%%/" -e "s/locavit,/locavit --/" <../auto-simplification/theirs.blob | tr % "\012" >theirs.blob + ) + mkdir conflict-without-LF + (cd conflict-without-LF + printf "line1\nline2\nline3" >base.blob + printf "line1\nline2\nline3x" >ours.blob + printf "line1\nline2\nline3y" >theirs.blob + ) + + mkdir marker-newline-handling-crlf + (cd marker-newline-handling-crlf + printf "1\\r\\n2\\r\\n3" >base.blob + printf "1\\r\\n2\\r\\n4" >ours.blob + printf "1\\r\\n2\\r\\n5" >theirs.blob + ) + + mkdir marker-newline-handling-lf + (cd marker-newline-handling-lf + printf "1\\r\\n2\\n3" >base.blob + printf "1\\r\\n2\\n4" >ours.blob + printf "1\\r\\n2\\n5" >theirs.blob + ) + + mkdir marker-newline-handling-lf2 + (cd marker-newline-handling-lf2 + printf "1\\r\\n2\\r\\n3" >base.blob + printf "1\\r\\n2\\n4" >ours.blob + printf "1\\r\\n2\\n5" >theirs.blob + ) + + mkdir spurious-c-conflicts + (cd spurious-c-conflicts + cp ../base.c base.blob + cp ../ours.c ours.blob + cp ../theirs.c theirs.blob + ) +) + + +for dir in simple \ + multi-change \ + clear-ours \ + clear-theirs \ + ours-2-lines-theirs-1-line \ + partial-match \ + unique-merge-base-with-insertion \ + zdiff3-basic \ + zdiff3-middlecommon \ + zdiff3-interesting \ + zdiff3-evil \ + no-change-add \ + no-change-remove \ + complex/no-change \ + complex/no-conflict \ + complex/no-conflict-too \ + complex/they-changed \ + complex/missing-LF-at-EOF \ + complex/missing-LF-at-EOF-no-conflict \ + complex/with-conflicts \ + complex/with-conflicts-in-removed-tail \ + complex/auto-simplification \ + complex/auto-simplification2 \ + complex/conflict-without-LF \ + complex/marker-newline-handling-crlf \ + complex/marker-newline-handling-lf \ + complex/marker-newline-handling-lf2 \ + complex/spurious-c-conflicts; do + DIR=$dir + baseline ours base theirs merge + baseline ours base theirs diff3 --diff3 + baseline ours base theirs zdiff3 --zdiff3 + baseline ours base theirs merge-ours --ours + baseline ours base theirs merge-theirs --theirs + baseline ours base theirs merge-union --union + ( + export GIT_CONFIG_COUNT=1 GIT_CONFIG_KEY_0=diff.algorithm GIT_CONFIG_KEY_VALUE=0=histogram + baseline ours base theirs diff3-histogram --diff3 + baseline ours base theirs zdiff3-histogram --zdiff3 + ) +done \ No newline at end of file diff --git a/gix-merge/tests/merge/blob/builtin_driver.rs b/gix-merge/tests/merge/blob/builtin_driver.rs new file mode 100644 index 00000000000..17901d7ff59 --- /dev/null +++ b/gix-merge/tests/merge/blob/builtin_driver.rs @@ -0,0 +1,202 @@ +use gix_merge::blob::builtin_driver::binary::{Pick, ResolveWith}; +use gix_merge::blob::{builtin_driver, Resolution}; + +#[test] +fn binary() { + assert_eq!( + builtin_driver::binary(None), + (Pick::Ours, Resolution::Conflict), + "by default it picks ours and marks it as conflict" + ); + assert_eq!( + builtin_driver::binary(Some(ResolveWith::Ancestor)), + (Pick::Ancestor, Resolution::Complete), + "Otherwise we can pick anything and it will mark it as complete" + ); + assert_eq!( + builtin_driver::binary(Some(ResolveWith::Ours)), + (Pick::Ours, Resolution::Complete) + ); + assert_eq!( + builtin_driver::binary(Some(ResolveWith::Theirs)), + (Pick::Theirs, Resolution::Complete) + ); +} + +mod text { + use bstr::ByteSlice; + use gix_merge::blob::Resolution; + use pretty_assertions::assert_str_eq; + + #[test] + fn run_baseline() -> crate::Result { + let root = gix_testtools::scripted_fixture_read_only("text-baseline.sh")?; + let cases = std::fs::read_to_string(root.join("baseline.cases"))?; + let mut out = Vec::new(); + for case in baseline::Expectations::new(&root, &cases) + // TODO: fix all of these eventually + .filter(|case| { + // Somehow, on in zdiff mode, it's different, and I wasn't able to figure out the rule properly. + // Now we prefer ancestor/before newlines and somewhat ignore our hunks. It's probably a minor issue in practice. + // gix: "1\r\n2\n<<<<<<< complex/marker-newline-handling-lf2/ours.blob\n4\r\n||||||| complex/marker-newline-handling-lf2/base.blob\r\n2\r\n3\n=======\n5\n>>>>>>> complex/marker-newline-handling-lf2/theirs.blob\n" + // git: "1\r\n2\n<<<<<<< complex/marker-newline-handling-lf2/ours.blob\n4 \n||||||| complex/marker-newline-handling-lf2/base.blob \n2\r\n3\n=======\n5\n>>>>>>> complex/marker-newline-handling-lf2/theirs.blob\n" + case.name != "complex/marker-newline-handling-lf2/zdiff3.merged" + && case.name != "complex/marker-newline-handling-lf2/zdiff3-histogram.merged" + }) + .filter(|case| { + // This is related to Git seemingly extending a hunk to increase overlap (see diff3) + case.name != "zdiff3-interesting/merge.merged" + && case.name != "zdiff3-interesting/merge-ours.merged" + && case.name != "zdiff3-interesting/diff3.merged" + && case.name != "zdiff3-interesting/diff3-histogram.merged" + && case.name != "zdiff3-interesting/zdiff3.merged" + && case.name != "zdiff3-interesting/zdiff3-histogram.merged" + && case.name != "zdiff3-interesting/merge-union.merged" + }) + .filter(|case| + // Git can extend hunks, similar to above, but the effect is not as noticeable. + // Implementing this would be interesting, to figure out when the hunk processing should apply. + case.name != "zdiff3-evil/merge.merged" && case.name != "zdiff3-evil/merge-union.merged") + .filter(|case| { + // Git seems to merge to hunks if they are close together to get a less noisy diff. + case.name != "zdiff3-middlecommon/merge.merged" && case.name != "zdiff3-middlecommon/merge-union.merged" + }) + .filter(|case| { + // Git has special character handling, which does magic to prevent conflicts + case.name != "complex/auto-simplification/merge.merged" + && case.name != "complex/auto-simplification/merge-union.merged" + }) + .filter(|case| { + // Git has special newline handling when diffing, + // which auto-inserts a newline when it was removed, kind of. + case.name != "complex/missing-LF-at-EOF/merge.merged" + && case.name != "complex/missing-LF-at-EOF/diff3.merged" + && case.name != "complex/missing-LF-at-EOF/diff3-histogram.merged" + && case.name != "complex/missing-LF-at-EOF/zdiff3.merged" + && case.name != "complex/missing-LF-at-EOF/zdiff3-histogram.merged" + && case.name != "complex/missing-LF-at-EOF/merge-ours.merged" + && case.name != "complex/missing-LF-at-EOF/merge-theirs.merged" + && case.name != "complex/missing-LF-at-EOF/merge-union.merged" + }) + .filter(|case| { + // Git has different diff-slider-heuristics so diffs can be different. + // See https://github.com/mhagger/diff-slider-tools. + case.name != "complex/spurious-c-conflicts/merge.merged" + && case.name != "complex/spurious-c-conflicts/merge-union.merged" + && case.name != "complex/spurious-c-conflicts/diff3-histogram.merged" + && case.name != "complex/spurious-c-conflicts/zdiff3-histogram.merged" + }) + { + let mut input = imara_diff::intern::InternedInput::default(); + dbg!(&case.name, case.options); + let actual = gix_merge::blob::builtin_driver::text( + &mut out, + &mut input, + &case.ours, + Some(case.ours_marker.as_str().as_ref()), + &case.base, + Some(case.base_marker.as_str().as_ref()), + &case.theirs, + Some(case.theirs_marker.as_str().as_ref()), + case.options, + ); + let expected_resolution = if case.expected.contains_str("<<<<<<<") { + Resolution::Conflict + } else { + Resolution::Complete + }; + assert_eq!(out.as_bstr(), case.expected); + assert_str_eq!( + out.as_bstr().to_str_lossy(), + case.expected.to_str_lossy(), + "{}: output mismatch\n{}", + case.name, + out.as_bstr() + ); + assert_eq!(actual, expected_resolution, "{}: resolution mismatch", case.name,); + } + Ok(()) + } + + mod baseline { + use bstr::BString; + use gix_merge::blob::builtin_driver::text::{ConflictStyle, ResolveWith}; + use std::path::Path; + + #[derive(Debug)] + pub struct Expectation { + pub ours: BString, + pub ours_marker: String, + pub theirs: BString, + pub theirs_marker: String, + pub base: BString, + pub base_marker: String, + pub name: BString, + pub expected: BString, + pub options: gix_merge::blob::builtin_driver::text::Options, + } + + pub struct Expectations<'a> { + root: &'a Path, + lines: std::str::Lines<'a>, + } + + impl<'a> Expectations<'a> { + pub fn new(root: &'a Path, cases: &'a str) -> Self { + Expectations { + root, + lines: cases.lines(), + } + } + } + + impl Iterator for Expectations<'_> { + type Item = Expectation; + + fn next(&mut self) -> Option { + let line = self.lines.next()?; + let mut words = line.split(' '); + let (Some(ours), Some(base), Some(theirs), Some(output)) = + (words.next(), words.next(), words.next(), words.next()) + else { + panic!("need at least the input and output") + }; + + let read = |rela_path: &str| read_blob(self.root, rela_path); + + let mut options = gix_merge::blob::builtin_driver::text::Options::default(); + for arg in words { + match arg { + "--diff3" => options.conflict_style = ConflictStyle::Diff3, + "--zdiff3" => options.conflict_style = ConflictStyle::ZealousDiff3, + "--ours" => options.on_conflict = Some(ResolveWith::Ours), + "--theirs" => options.on_conflict = Some(ResolveWith::Theirs), + "--union" => options.on_conflict = Some(ResolveWith::Union), + _ => panic!("Unknown argument to parse into options: '{arg}'"), + } + } + if output.contains("histogram") { + options.diff_algorithm = imara_diff::Algorithm::Histogram; + } + + Some(Expectation { + ours: read(ours), + ours_marker: ours.into(), + theirs: read(theirs), + theirs_marker: theirs.into(), + base: read(base), + base_marker: base.into(), + expected: read(output), + name: output.into(), + options, + }) + } + } + + fn read_blob(root: &Path, rela_path: &str) -> BString { + std::fs::read(root.join(rela_path)) + .unwrap_or_else(|_| panic!("Failed to read '{rela_path}' in '{}'", root.display())) + .into() + } + } +} diff --git a/gix-merge/tests/merge/blob/mod.rs b/gix-merge/tests/merge/blob/mod.rs new file mode 100644 index 00000000000..f781f63e485 --- /dev/null +++ b/gix-merge/tests/merge/blob/mod.rs @@ -0,0 +1 @@ +mod builtin_driver; diff --git a/gix-merge/tests/merge/main.rs b/gix-merge/tests/merge/main.rs new file mode 100644 index 00000000000..05375cb2279 --- /dev/null +++ b/gix-merge/tests/merge/main.rs @@ -0,0 +1,4 @@ +#[cfg(feature = "blob")] +mod blob; + +pub use gix_testtools::Result;