* Use buffered stdout to reduce write sys calls.
This simple change yielded the biggest performace gain.
* Use `for_byte_record_with_terminator` from the `bstr` crate.
This is to minimize the per line copying needed by
`BufReader::read_until`. The `cut_fields` and `cut_fields_delimiter`
functions used `read_until` to iterate over lines. That required copying
each input line to the line buffer. With
`for_byte_record_with_terminator` copying is minimized as it calls our
closure with a reference to BufReader's buffer most of the time. It
needs to copy (internally) only to process any incomplete lines at the
end of the buffer.
* Re-write `Searcher` to use `memchr`.
Switch from the naive implementation to one that uses `memchr`.
* Rewrite `cut_bytes` almost entirely.
This was already well optimized. The performance gain in this case is
not from avoiding copying. In fact, it needed zero copying whereas new
implementation introduces some copying similar to `cut_fields` described
above. But the occassional copying cost is more than offset by the use
of the very fast `memchr` inside `for_byte_record_with_terminator`.
This change also simplifies the code significantly. Removed the `buffer`
module.
This adds a --debug flag, which, when activated, will draw lines below
the characters that are actually used for comparisons.
This is not a complete implementation of --debug. It should, quoting the man page
for GNU sort: "annotate the part of the line used to sort, and warn
about questionable usage to stderr". Warning about "questionable usage"
is not part of this patch.
This change required some adjustments to be able to get the range that
is actually used for comparisons. Most notably, general numeric comparisons
were rewritten, fixing some bugs along the lines.
Testing is mostly done by adding fixtures for the expected debug output of
existing tests.
* ls: Remove allocations by eliminating collect/clones
* ls: Introduce PathData structure
- PathData will hold Path related metadata / strings that are required
frequently in subsequent functions
- All data is precomputed and cached and subsequent functions just
use cached data
* ls: Cache more data related to paths
- Cache filename and sort by filename instead of full path
- Cache uid->usr and gid->grp mappings
https://github.com/uutils/coreutils/pull/2099/files
* ls: Add BENCHMARKING.md
* ls: Document PathData structure
* tests/ls: Add testcase for error paths with width option
* ls: Fix unused import warning
cached will be only used for unix currently as current use of
caching gid/uid mappings is only relevant on unix
* ls: Suggest checking syscall count in BENCHMARKING.md
* ls: Remove mentions of sort in BENCHMARKING.md
* ls: Remove dependency on cached
Implement caching using HashMap and lazy_static
* ls: Fix MSRV error related to map_or
Rust 1.40 did not support map_or for result types
Trailing separators were included at the end of the last token, but they
should not be.
This changes tokenize_with_separator as suggested by @cbjadwani.
GNU sort disallows these combinations, presumably because they are
likely not what the user really wants.
Ignoring characters would cause things to be put together that aren't
together in the input. For example, -dn would cause "0.12" or "0,12" to
be parsed as "12" which is highly unexpected and confusing.
This reduces memory usage by only storing two lines of the input file at
a time. The current implementation first builds a list of all duplicate
lines ('group') and then decides which lines of the group should be
printed.
- Passing `never` to `--reflink` does not raise an error anymore.
- Remove `Options::reflink` flag as it was redundant with
`reflink_mode`.
- Add basic tests for this option. Does not check that a copy-on-write
rather than a regular copy was made.
* sort: use unstable sort when possible
This results in a very minor performance (speed) improvement.
It does however result in a memory usage reduction, because unstable
sort does not allocate auxiliary memory. There's also an improvement in
overall CPU usage.
* add benchmarking instructions
* add user time
* fix typo
* sort: implement numeric string comparison
This implements -n and -h using a string comparison algorithm instead
of parsing each number to a f64 and comparing those.
This should result in a moderate performance increase and eliminate loss
of precision.
* cache parsed f64 numbers
For general numeric comparisons we have to parse numbers as f64,
as this behavior is explicitly documented by GNU coreutils.
We can however cache the parsed value to speed up comparisons.
* fix leading zeroes for negative numbers
* use more appropriate name for exponent
* improvements to the parse function
* move checks into main loop and fix thousands separator condition
* remove unneeded checks
* rustfmt
* du error output should match GNU
* Created a new error macro which allows the customization of the
"error:" string part
* Match the du output based on the type of error encountered. Can extend
to handling other errors I guess.
* Rustfmt updates
* Added non-windows test for du no permission output
