Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 94df6e0

Browse files
ByronByron
committed
feat!: a complete, streaming compatible, git filter pipeline.
It's configuring itself based on each input path and supports streaming. Some name changes where also performed while at it.
1 parent 8a361b4 commit 94df6e0

25 files changed

Lines changed: 1082 additions & 66 deletions

File tree

‎Cargo.lock‎

Lines changed: 3 additions & 0 deletions
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

‎gix-filter/Cargo.toml‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -19,12 +19,15 @@ gix-command = { version = "^0.2.6", path = "../gix-command" }
1919
gix-quote = { version = "^0.4.5", path = "../gix-quote" }
2020
gix-path = { version = "^0.8.3", path = "../gix-path" }
2121
gix-packetline = { package = "gix-packetline-blocking", version = "^0.16.3", path = "../gix-packetline-blocking" }
22+
gix-attributes = { version = "0.14.1", path = "../gix-attributes" }
2223

2324
encoding_rs = "0.8.32"
2425
bstr = { version = "1.5.0", default-features = false, features = ["std"] }
2526
thiserror = "1.0.38"
27+
smallvec = "1.10.0"
2628

2729

2830
[dev-dependencies]
2931
once_cell = "1.18.0"
3032
gix-testtools = { path = "../tests/tools" }
33+
gix-worktree = { version = "0.21.1", path = "../gix-worktree" }

gix-filter/examples/ident.rs renamed to gix-filter/examples/arrow.rs

File renamed without changes.

‎gix-filter/src/driver/apply.rs‎

Lines changed: 10 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -40,11 +40,11 @@ pub enum Error {
4040

4141
/// Additional information for use in the [`State::apply()`] method.
4242
#[derive(Debug, Copy, Clone)]
43-
pub struct Context<'a> {
43+
pub struct Context<'a, 'b> {
4444
/// The repo-relative using slashes as separator of the entry currently being processed.
4545
pub rela_path: &'a BStr,
4646
/// The name of the reference that `HEAD` is pointing to. It's passed to `process` filters if present.
47-
pub ref_name: Option<&'a BStr>,
47+
pub ref_name: Option<&'b BStr>,
4848
/// The root-level tree that contains the current entry directly or indirectly, or the commit owning the tree (if available).
4949
///
5050
/// This is passed to `process` filters if present.
@@ -58,6 +58,7 @@ pub struct Context<'a> {
5858
impl State {
5959
/// Apply `operation` of `driver` to the bytes read from `src` and return a reader to immediately consume the output
6060
/// produced by the filter. `rela_path` is the repo-relative path of the entry to handle.
61+
/// It's possible that the filter stays inactive, in which case the `src` isn't consumed and has to be used by the caller.
6162
///
6263
/// Each call to this method will cause the corresponding filter to be invoked unless `driver` indicates a `process` filter,
6364
/// which is only launched once and maintained using this state.
@@ -74,9 +75,9 @@ impl State {
7475
pub fn apply<'a>(
7576
&'a mut self,
7677
driver: &Driver,
77-
src: impl std::io::Read,
78+
src: &mut impl std::io::Read,
7879
operation: Operation,
79-
ctx: Context<'_>,
80+
ctx: Context<'_, '_>,
8081
) -> Result<Option<Box<dyn std::io::Read + 'a>>, Error> {
8182
match self.apply_delayed(driver, src, operation, Delay::Forbid, ctx)? {
8283
Some(MaybeDelayed::Delayed(_)) => {
@@ -94,14 +95,14 @@ impl State {
9495
pub fn apply_delayed<'a>(
9596
&'a mut self,
9697
driver: &Driver,
97-
mut src: impl std::io::Read,
98+
src: &mut impl std::io::Read,
9899
operation: Operation,
99100
delay: Delay,
100-
ctx: Context<'_>,
101+
ctx: Context<'_, '_>,
101102
) -> Result<Option<MaybeDelayed<'a>>, Error> {
102-
match self.process(driver, operation, ctx.rela_path)? {
103+
match self.maybe_launch_process(driver, operation, ctx.rela_path)? {
103104
Some(Process::SingleFile { mut child, command }) => {
104-
std::io::copy(&mut src, &mut child.stdin.take().expect("configured"))?;
105+
std::io::copy(src, &mut child.stdin.take().expect("configured"))?;
105106
Ok(Some(MaybeDelayed::Immediate(Box::new(ReadFilterOutput {
106107
inner: child.stdout.take(),
107108
child: driver.required.then_some((child, command)),
@@ -179,7 +180,7 @@ impl State {
179180
}
180181
}
181182

182-
/// A utility type to represent delayed or immediate apply-filter results.
183+
/// A type to represent delayed or immediate apply-filter results.
183184
pub enum MaybeDelayed<'a> {
184185
/// Using the delayed protocol, this entry has been sent to a long-running process and needs to be
185186
/// checked for again, later, using the [`driver::Key`] to refer to the filter who owes a response.

‎gix-filter/src/driver/init.rs‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,7 @@ use crate::{driver, Driver};
33
use bstr::{BStr, BString};
44
use std::process::Stdio;
55

6-
/// The error returned by [State::process()][super::State::process()].
6+
/// The error returned by [State::maybe_launch_process()][super::State::maybe_launch_process()].
77
#[derive(Debug, thiserror::Error)]
88
#[allow(missing_docs)]
99
pub enum Error {
@@ -24,7 +24,7 @@ impl State {
2424
/// file for use in the invoked `SingleFile` process.
2525
///
2626
/// Note that if a long-running process is defined, the `operation` isn't relevant and capabilities are to be checked by the caller.
27-
pub fn process(
27+
pub fn maybe_launch_process(
2828
&mut self,
2929
driver: &Driver,
3030
operation: Operation,

‎gix-filter/src/driver/shutdown.rs‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -11,7 +11,7 @@ pub enum Mode {
1111
Ignore,
1212
}
1313

14-
/// Shutdown
14+
/// Lifecycle
1515
impl State {
1616
/// Handle long-running processes according to `mode`. If an error occours, all remaining processes will be ignored automatically.
1717
/// Return a list of `(process, Option<status>)`

‎gix-filter/src/eol/mod.rs‎

Lines changed: 33 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -18,6 +18,18 @@ pub enum Mode {
1818
CrLf,
1919
}
2020

21+
/// Possible states for the `core.autocrlf`.
22+
#[derive(Default, Debug, Copy, Clone, Eq, PartialEq)]
23+
pub enum AutoCrlf {
24+
/// The same as if the `text eol=lf` attribute is set.
25+
Input,
26+
/// The same as if the `text eol=crlf` attribute is set.
27+
Enabled,
28+
/// No conversion is performed.
29+
#[default]
30+
Disabled,
31+
}
32+
2133
/// The combination of `crlf`, `text` and `eol` attributes into one neat package.
2234
#[derive(Debug, Copy, Clone, Eq, PartialEq)]
2335
pub enum AttributesDigest {
@@ -37,11 +49,30 @@ pub enum AttributesDigest {
3749
TextAutoInput,
3850
}
3951

52+
impl From<Mode> for AttributesDigest {
53+
fn from(value: Mode) -> Self {
54+
match value {
55+
Mode::Lf => AttributesDigest::TextInput,
56+
Mode::CrLf => AttributesDigest::TextCrlf,
57+
}
58+
}
59+
}
60+
61+
impl From<AutoCrlf> for AttributesDigest {
62+
fn from(value: AutoCrlf) -> Self {
63+
match value {
64+
AutoCrlf::Input => AttributesDigest::TextAutoInput,
65+
AutoCrlf::Enabled => AttributesDigest::TextAutoCrlf,
66+
AutoCrlf::Disabled => AttributesDigest::Binary,
67+
}
68+
}
69+
}
70+
4071
/// Git Configuration that affects how CRLF conversions are applied.
4172
#[derive(Default, Debug, Copy, Clone)]
4273
pub struct Configuration {
43-
/// Corresponds to `core.autocrlf` and is `None` for `input`, `Some(true)` if `true` or `Some(false)` if `false`.
44-
pub auto_crlf: Option<bool>,
74+
/// Corresponds to `core.autocrlf`.
75+
pub auto_crlf: AutoCrlf,
4576
/// Corresponds to `core.eol`, and is `None` if unset or set to `native`, or `Some(<mode>)` respectively.
4677
pub eol: Option<Mode>,
4778
}

‎gix-filter/src/eol/utils.rs‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
use crate::eol::{AttributesDigest, Configuration, Mode, Stats};
1+
use crate::eol::{AttributesDigest, AutoCrlf, Configuration, Mode, Stats};
22

33
impl Default for Mode {
44
fn default() -> Self {
@@ -34,9 +34,9 @@ impl Configuration {
3434
/// Return the line-ending mode that is configured here.
3535
pub fn to_eol(&self) -> Mode {
3636
match self.auto_crlf {
37-
Some(true) => Mode::CrLf,
38-
None => Mode::Lf,
39-
Some(false) => self.eol.unwrap_or_default(),
37+
AutoCrlf::Enabled => Mode::CrLf,
38+
AutoCrlf::Input => Mode::Lf,
39+
AutoCrlf::Disabled => self.eol.unwrap_or_default(),
4040
}
4141
}
4242
}

‎gix-filter/src/lib.rs‎

Lines changed: 34 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -12,18 +12,48 @@
1212

1313
use bstr::BString;
1414

15-
///
15+
/// A forwarding of the `encoding_rs` crate for its types and convenience.
16+
pub use encoding_rs as encoding;
17+
18+
/// a filter to replace `$Id$` with a git-hash of the buffer.
1619
pub mod ident;
1720

18-
/// utilities related to handling line endings in buffers
21+
/// convert line endings in buffers
1922
pub mod eol;
2023

21-
/// Utilities for handling worktree encodings.
24+
/// change encodings based on the `working-tree-encoding` attribute.
2225
pub mod worktree;
2326

24-
/// Utilities around driver programs.
27+
/// use filter programs to perform any kind of conversion.
2528
pub mod driver;
2629

30+
///
31+
pub mod pipeline;
32+
33+
/// The standard git filter pipeline comprised of multiple standard filters and support for external filters.
34+
///
35+
/// It's configuring itself for each provided path based on the path's attributes, implementing the complex logic that governs it.
36+
pub struct Pipeline {
37+
/// all available filter programs as configured in git-config, for use with `filter=driver` attributes.
38+
drivers: Vec<Driver>,
39+
/// Storage for the attributes of each item we should process, configured for use with all attributes that concern us.
40+
attrs: gix_attributes::search::Outcome,
41+
/// configuration sourced from git-config to decide how to deal with the eol filter.
42+
eol_config: eol::Configuration,
43+
/// A list of encodings for which round-trip checks should be performed when changing the worktree encoding to git.
44+
encodings_with_roundtrip_check: Vec<&'static encoding_rs::Encoding>,
45+
/// Whether to check for roundtrip safety during CRLF encodings, and what to do on roundtrip failure
46+
crlf_roundtrip_check: pipeline::CrlfRoundTripCheck,
47+
/// Additional context to pass to process filters.
48+
context: pipeline::Context,
49+
/// State needed to keep running filter processes.
50+
processes: driver::State,
51+
/// The kind of object hashes to create for the `ident` filter.
52+
object_hash: gix_hash::Kind,
53+
/// A utility to handle multiple buffers to keep results of various filters.
54+
bufs: pipeline::util::Buffers,
55+
}
56+
2757
/// A declaration of a driver program.
2858
///
2959
/// It consists of up to three program declarations.

0 commit comments

Comments
 (0)