This brings in a bug fix that no longer tries to run `git` to update the
submodule if the `git` command doesn't exist.
This is useful is more restricted build contexts where `git` isn't
installed. Such as in the docker image used for running `cross`.
It turns out that musl's allocator is slow enough to cause a fairly
noticeable performance regression when ripgrep is built as a static
binary with musl. We fix this by using jemalloc when building with musl.
We continue to use the default system allocator in all other scenarios.
Namely, glibc's allocator doesn't noticeably regress performance compared
to jemalloc. But we could add more targets to this logic if other
system allocators (macOS, Windows) prove to be slow.
This wasn't necessary before because rustc recently stopped using jemalloc
by default.
Fixes#1268
This makes the case of searching for a dictionary of a very large number
of literals much much faster. (~10x or so.) In particular, we achieve this
by short-circuiting the construction of a full regex when we know we have
a simple alternation of literals. Building the regex for a large dictionary
(>100,000 literals) turns out to be quite slow, even if it internally will
dispatch to Aho-Corasick.
Even that isn't quite enough. It turns out that even *parsing* such a regex
is quite slow. So when the -F/--fixed-strings flag is set, we short
circuit regex parsing completely and jump straight to Aho-Corasick.
We aren't quite as fast as GNU grep here, but it's much closer (less than
2x slower).
In general, this is somewhat of a hack. In particular, it seems plausible
that this optimization could be implemented entirely in the regex engine.
Unfortunately, the regex engine's internals are just not amenable to this
at all, so it would require a larger refactoring effort. For now, it's
good enough to add this fairly simple hack at a higher level.
Unfortunately, if you don't pass -F/--fixed-strings, then ripgrep will
be slower, because of the aforementioned missing optimization. Moreover,
passing flags like `-i` or `-S` will cause ripgrep to abandon this
optimization and fall back to something potentially much slower. Again,
this fix really needs to happen inside the regex engine, although we
might be able to special case -i when the input literals are pure ASCII
via Aho-Corasick's `ascii_case_insensitive`.
Fixes#497, Fixes#838
This commit adds a new encoding feature where the -E/--encoding flag
will now accept a value of 'none'. When given this value, all encoding
related machinery is disabled and ripgrep will search the raw bytes of
the file, including the BOM if it's present.
Closes#1207, Closes#1208
This undoes the patch to stop using bytecount on big-endian
architectures. In particular, we bump our bytecount dependency to the
latest release, which has a fix.
This reverts commit a4868b8835.
Fixes#1144 (again), Closes#1194
This brings in an updated `encoding_rs` crate that uses `packed_simd`,
which compiles on the latest nightly. Compilation times do appear to be
impacted significantly though.
Fixes#1175 (again)
This commit fixes a bug where ripgrep only treated files beginning with
a `.` as hidden. On Windows, we continue this tradition, but
additionally check whether a file has the special Windows "hidden"
attribute set. If so, we treat it as a hidden file.
In order to make this work without an additional stat call, we had to
rearrange some of the plumbing from the directory traverser.
Fixes#1154
This fixes a bug where a BOM prefix was included. While this was somewhat
intentional in order to have a faithful "UTF8 passthru" option, in
practice, this causes problems such as breaking patterns like `^` in a
really non-obvious way.
The actual fix was to add a new API to encoding_rs_io, which this commit
brings in.
Fixes#1163
bytecount now uses runtime dispatch for enabling SIMD, which means we can
no longer need the avx-accel features. We remove it from ripgrep since the
next release will be a minor version bump, but leave them as no-ops for
the crates that previously used it.
This commit is the result of doing:
$ cargo update
$ cargo update -p encoding_rs --precise 0.8.10
where the latter line prevents encoding_rs from updating to 0.8.11 (or
newer). In particular, the 0.8.11 release increased the minimum Rust
version to 1.29, where as ripgrep 0.10.x is still on 1.28. We stay on an
older version for now until ripgrep is ready to move to 0.11.x.
This also requires corresponding updates to both rand and rand_core. Doing
an update of rand without doing an update of rand_core results in
compilation errors because two distinct versions of rand_core are included
in the build, and the traits they expose are distinct and incompatible.
We also switch over to using tempfile instead of tempdir, which drops the
last remaining thing keeping rand 0.4 in the build.
Fixes#1141, Fixes#1142
This brings in some new Unicode properties, such as \p{Emoji}.
It is now also technically possible construct a regex that recognizes
grapheme clusters.
This commit moves a lot of "utility" code from ripgrep core into
grep-cli. Any one of these things might not be worth creating a new
crate, but combining everything together results in a fair number of a
convenience routines that make up a decent sized crate.
There is potentially more we could move into the crate, but much of what
remains in ripgrep core is almost entirely dealing with the number of
flags we support.
In the course of doing moving things to the grep-cli crate, we clean up
a lot of gunk and improve failure modes in a number of cases. In
particular, we've fixed a bug where other processes could deadlock if
they write too much to stderr.
Fixes#990
This commit adds a 'same_file_system' option to the walk builder. For
single threaded walking, it defers to the walkdir crate, which has the
same option. The bulk of this commit implements this flag for the parallel
walker. We add one very feeble test for this.
The parallel walker is now officially a complete mess.
Closes#321
This commit fixes a bug where the first path always reported itself as
as symlink via `path_is_symlink`.
Part of this fix includes updating walkdir to 2.2.1, which also includes
a corresponding bug fix.
Fixes#984
This also updates some code to make use of our more liberal versioning
requirement, including the use of crossbeam-channel instead of the MsQueue
from the older an unmaintained crossbeam 0.3. This does regrettably add
a sizable number of dependencies, however, compile times seem mostly
unaffected.
Closes#1019
This basically rewrites every integration test. We reduce the amount of
magic involved here in terms of which arguments are being passed to
ripgrep processes. To make up for the boiler plate saved by the magic,
we make the Dir (formerly WorkDir) type a bit nicer to use, along with a
new TestCommand that wraps a std::process::Command. In exchange, we get
tests that are easier to read and write.
We also run every test with the `--pcre2` flag to make sure that works,
when PCRE2 is available.
This commit does the work to delete the old `grep` crate and effectively
rewrite most of ripgrep core to use the new libripgrep crates. The new
`grep` crate is now a facade that collects the various crates that make
up libripgrep.
The most complex part of ripgrep core is now arguably the translation
between command line parameters and the library options, which is
ultimately where we want to be.
libripgrep is not any one library, but rather, a collection of libraries
that roughly separate the following key distinct phases in a grep
implementation:
1. Pattern matching (e.g., by a regex engine).
2. Searching a file using a pattern matcher.
3. Printing results.
Ultimately, both (1) and (3) are defined by de-coupled interfaces, of
which there may be multiple implementations. Namely, (1) is satisfied by
the `Matcher` trait in the `grep-matcher` crate and (3) is satisfied by
the `Sink` trait in the `grep2` crate. The searcher (2) ties everything
together and finds results using a matcher and reports those results
using a `Sink` implementation.
Closes#162
winapi 0.3.5 changed how it represents some of its structs, which caused
a bug to surface in atty that prevents tty detection on Windows. atty
has an open PR to fix this: https://github.com/softprops/atty/pull/28
Until a new release of atty, we pin winapi to a version that works.
This commit mostly moves the transcoder implementation to its own
crate: https://github.com/BurntSushi/encoding_rs_io
The new crate adds clear documentation and cleans up the implementation
to fully implement the contract of io::Read.
This causes SIMD to kick in automatically when compiling with stable
Rust 1.27+.
We also update the README to describe the current state of things.
Thanks to @hartley for pointing this out:
https://twitter.com/hartley/status/1009950392862453760
atty 0.2.7 (and 0.2.8) contain a regression in cygwin terminals that
prevents basic use of ripgrep, and is also the cause of the Windows CI
test failures. For now, we pin to 0.2.6, but a patch has been submitted
upstream: https://github.com/softprops/atty/pull/25
Nothing to see here.
Note that we continue to refrain to update tempdir, which means we are
still bringing in rand 0.4 and rand 0.3. Updating tempdir brings in an
old version of remove_dir_all, which in turn brings in winapi 0.2. No
thanks.
This update brings with it many bug fixes:
* Better error messages are printed overall. We also include
explicit call out for unsupported features like backreferences
and look-around.
* Regexes like `\s*{` no longer emit incomprehensible errors.
* Unicode escape sequences, such as `\u{..}` are now supported.
For the most part, this upgrade was done in a straight-forward way. We
resist the urge to refactor the `grep` crate, in anticipation of it
being rewritten anyway.
Note that we removed the `--fixed-strings` suggestion whenever a regex
syntax error occurs. In practice, I've found that it results in a lot of
false positives, and I believe that its use is not as paramount now that
regex parse errors are much more readable.
Closes#268, Closes#395, Closes#702, Closes#853
This update brings with it a new feature of the regex crate which will
now use SIMD optimizations automatically at runtime with no necessary
compile time flags. All that's needed is to enable the `unstable` feature.
Other crates, such as bytecount and encoding_rs, are still using the
old-style SIMD support, so we leave the simd-accel and avx-accel features.
However, the binaries we distribute on Github no longer have those
features enabled, which makes them truly portable.
Fixes#135
This commit fixes a performance regression in Windows that resulted from
fallout from fixing #705. In particular, we introduced an additional
stat call for every single directory entry, which can be quite
disastrous for performance.
There is a corresponding companion PR that fixes the same bug in
walkdir: https://github.com/BurntSushi/walkdir/pull/96Fixes#820
This regex update disabled the Tuned Boyer-Moore literal searcher which
has a bug in it that isn't straight-forward to fix. We bring that update
into ripgrep with this commit.
Fixes#780, Fixes#781
We use the new AppSettings::AllArgsOverrideSelf to permit all flags to
be specified multiple times. This removes the need for our previous
work-around where we would enable `multiple` for every flag and then
just extract the last value when consuming clap's matches.
We also add a couple regression tests that ensure repeated switches and
flags work as expected.
This removes the vec-map feature from clap. clap's README claims that
vec-map provides a small performance benefit, but I could observe any in
ripgrep workloads.
The benefit here is that it drops a dependency.
Amazingly, this drops whole release build times for ripgrep from 68s to
33s, and debug build time also decreases from 22s to 15.5s. This was
entirely unintentional but a welcome surprise.
This commit updates the `log` crate to 0.4 and drops the dependency on
env_logger. In particular, the latest version of env_logger brings in
additional non-optional dependencies such as chrono that I don't think is
worth including into ripgrep.
It turns out ripgrep doesn't need any fancy logging. We just need a concept
of log levels and the ability to print to stderr. Therefore, we just roll
our own super simple logger.
This update is motivated by the persistent configuration task. In
particular, we need the ability to toggle the global log level more than
once, and this doesn't appear to be possible with older versions of the
log crate.
This commit fixes a bug on Windows where directory traversals were
completely broken when attempting to scan OneDrive directories that use
the "file on demand" strategy.
The specific problem was that Rust's standard library treats OneDrive
directories as reparse points instead of directories, which causes
methods like `FileType::is_file` and `FileType::is_dir` to always return
false, even when retrieved via methods like `metadata` that purport to
follow symbolic links.
We fix this by peppering our code with checks on the underlying file
attributes exposed by Windows. We consider an entry a directory if and
only if the directory bit is set on the attributes. We are careful to
make sure that the code remains the same on non-Windows platforms.
Note that we also bump the dependency on `walkdir`, which contains a
similar fix for its traversals.
This bug is recorded upstream:
https://github.com/rust-lang/rust/issues/46484
Upstream also has a pending PR:
https://github.com/rust-lang/rust/pull/47956Fixes#705
This commit updates to the latest walkdir release, which fixes a bug on
Windows where ripgrep would panic if it was told to traverse a directory
while following symlinks *and* if opening one of those symlinks failed.
Fixes#633
This commit adds opt-in support for searching compressed files during
recursive search. This behavior is only enabled when the
`-z/--search-zip` flag is passed to ripgrep. When enabled, a limited set
of common compression formats are recognized via file extension, and a
new process is spawned to perform the decompression. ripgrep then
searches the stdout of that spawned process.
Closes#539
The regex update fixes the Rust nightly build failure by in turn updating
its simd dependency to 2.x.
The regex update also includes a literal optimization that uses Tuned
Boyer Moore.
Fixes#617
The uninteresting bits of this commit involve mechanical changes for
updates to walkdir 2. The more interesting bits of this commit are the
breaking changes, although none of them should require any significant
change on users of this library. The breaking changes are as follows:
* `DirEntry::path_is_symbolic_link` has been renamed to
`DirEntry::path_is_symlink`. This matches the conventions in the
standard library, and also the corresponding name change in walkdir.
* Removed the `From<walkdir::Error> for ignore::Error` impl. This was
intended to only be used internally, but was the only thing that
made `walkdir` a public dependency of `ignore`. Therefore, we remove
it since it seems unnecessary.
* Renamed `WalkBuilder::sort_by` to `WalkBuilder::sort_by_file_name`,
and changed the type of the comparator from
Fn(&OsString, &OsString) -> cmp::Ordering + 'static
to
Fn(&OsStr, &OsStr) -> cmp::Ordering + Send + Sync + 'static
The corresponding change in `walkdir` retains the `sort_by` name, but
gives the comparator a pair of `&DirEntry` values instead of a pair
of `&OsStr` values. Ideally, `ignore` would hand off its own pair of
`&ignore::DirEntry` values, but this requires more design work. So for
now, we retain previous functionality, but leave room to make a proper
`sort_by` method.
[breaking-change]
This only bumps the regex dependency. The new clap version causes a bump
in unicode-segmentation, which in turn requires a Rust 1.15, which is
above ripgrep's currently supported minimum Rust version of 1.12.
This commit updates clap to v2.23.0
The update contained a bug fix in clap that results in broken code in
ripgrep. ripgrep was relying on the bug, but this commit fixes that
issue. The bug centered around not being able to override the
auto-generated help message by supplying a flag with a long of `help`.
Normally, supplying a flag with a long of `help` means whenever the user
passes `--help`, the consuming code (e.g. ripgrep) is responsible for
displaying the help message. However, due to the bug in clap this wasn't
necessary for ripgrep to do unless the user passed `-h`. With the bug
fixed, it meant the user passing `--help` and clap expected ripgrep to
display the help, yet ripgrep expected clap to display the help. This
has been fixed in this commit of ripgrep.
All well now!
v2.23.0 also brings the abilty to use `Arg::help` or `Arg::long_help`
allowing one to distinguish between `-h` and `--help`. This commit
leaves all doc strings in the `lazy_static!` hashmap however only for
aesthetic reasons.
This means all home rolled handling of `-h`/`--help` has been removed
from ripgrep, yet functionality *and* appearances are 100% the same.
This includes, but is not limited to, UTF-16, latin-1, GBK, EUC-JP and
Shift_JIS. (Courtesy of the `encoding_rs` crate.)
Specifically, this feature enables ripgrep to search files that are
encoded in an encoding other than UTF-8. The list of available encodings
is tied directly to what the `encoding_rs` crate supports, which is in
turn tied to the Encoding Standard. The full list of available encodings
can be found here: https://encoding.spec.whatwg.org/#concept-encoding-get
This pull request also introduces the notion that text encodings can be
automatically detected on a best effort basis. Currently, the only
support for this is checking for a UTF-16 bom. In all other cases, a
text encoding of `auto` (the default) implies a UTF-8 or ASCII
compatible source encoding. When a text encoding is otherwise specified,
it is unconditionally used for all files searched.
Since ripgrep's regex engine is fundamentally built on top of UTF-8,
this feature works by transcoding the files to be searched from their
source encoding to UTF-8. This transcoding only happens when:
1. `auto` is specified and a non-UTF-8 encoding is detected.
2. A specific encoding is given by end users (including UTF-8).
When transcoding occurs, errors are handled by automatically inserting
the Unicode replacement character. In this case, ripgrep's output is
guaranteed to be valid UTF-8 (excluding non-UTF-8 file paths, if they
are printed).
In all other cases, the source text is searched directly, which implies
an assumption that it is at least ASCII compatible, but where UTF-8 is
most useful. In this scenario, encoding errors are not detected. In this
case, ripgrep's output will match the input exactly, byte-for-byte.
This design may not be optimal in all cases, but it has some advantages:
1. In the happy path ("UTF-8 everywhere") remains happy. I have not been
able to witness any performance regressions.
2. In the non-UTF-8 path, implementation complexity is kept relatively
low. The cost here is transcoding itself. A potentially superior
implementation might build decoding of any encoding into the regex
engine itself. In particular, the fundamental problem with
transcoding everything first is that literal optimizations are nearly
negated.
Future work should entail improving the user experience. For example, we
might want to auto-detect more text encodings. A more elaborate UX
experience might permit end users to specify multiple text encodings,
although this seems hard to pull off in an ergonomic way.
Fixes#1
When running ripgrep like this:
rg foo > output
we must be careful not to search `output` since ripgrep is actively writing
to it. Searching it can cause massive blowups where the file grows without
bound.
While this is conceptually easy to fix (check the inode of the redirection
and the inode of the file you're about to search), there are a few problems
with it.
First, inodes are a Unix thing, so we need a Windows specific solution to
this as well. To resolve this concern, I created a new crate, `same-file`,
which provides a cross platform abstraction.
Second, stat'ing every file is costly. This is not avoidable on Windows,
but on Unix, we can get the inode number directly from directory traversal.
However, this information wasn't exposed, but now it is (through both the
ignore and walkdir crates).
Fixes#286
This means that ripgrep will no longer try to reset your colors in your
terminal if you kill it while searching. This could result in messing up
the colors in your terminal, and the fix is to simply run some other
command that resets them for you. For example:
$ echo -ne "\033[0m"
The reason why the ^C handling was removed is because it is irrevocably
broken on Windows and is impossible to do correctly and efficiently in
ANSI terminals.
Fixes#281
When give an explicit file path on the command line like `foo` where `foo`
is a symlink, ripgrep should follow it even if `-L` isn't set. This is
consistent with the behavior of `foo/`.
Fixes#256
This commit completely guts all of the color handling code and replaces
most of it with two new crates: wincolor and termcolor. wincolor
provides a simple API to coloring using the Windows console and
termcolor provides a platform independent coloring API tuned for
multithreaded command line programs. This required a lot more
flexibility than what the `term` crate provided, so it was dropped.
We instead switch to writing ANSI escape sequences directly and ignore
the TERMINFO database.
In addition to fixing several bugs, this commit also permits end users
to customize colors to a certain extent. For example, this command will
set the match color to magenta and the line number background to yellow:
rg --colors 'match:fg:magenta' --colors 'line:bg:yellow' foo
For tty handling, we've adopted a hack from `git` to do tty detection in
MSYS/mintty terminals. As a result, ripgrep should get both color
detection and piping correct on Windows regardless of which terminal you
use.
Finally, switch to line buffering. Performance doesn't seem to be
impacted and it's an otherwise more user friendly option.
Fixes#37, Fixes#51, Fixes#94, Fixes#117, Fixes#182, Fixes#231
