Always memchr(line_delim) which is fast and allows:
- skipping whole segments when the next selected byte is beyond them
- skipping unselected prefixes in bulk
- writing contiguous selected spans in bulk
This wins for lines >= 4 characters,
but is slower lines <= 3 characters, especially if selecting bytes 1-3.
That is unusual though.
This is about 20x faster.
Note we only do the delimiter search once per chunk,
and it's usually quick as delimiters wouldn't be too far
into the a chunk if present, so we don't bother
to cache the found delimiter.
* src/cut.c (usage): Mention blank characters are used to separate.
* doc/coreutils.texi (cut invocation): Likewise. Also describe
the 'trimmed' argument and the relation to -F.
Allows better/simpler avoidance of repeated line/delim scans
TODO: speed up our really slow cut_fields_mb_any.
Compare for example:
time src/cut -w -f1 ll.in >/dev/null #14s
time src/cut -d, -f1 ll.in >/dev/null #.1s
Could adjust so that LC_ALL=C does memchr2(space,tab) ?
Not sure about this at all.
Only worthwhile if can also remove cut_fields_line_delim.
Refactored src/cut.c so the old UTF-8 byte-search path is now the
general byte-search field engine for all safe byte-search cases:
ordinary single-byte delimiters and valid UTF-8 delimiters. The old
cut_fields path did not go away completely; it is now
cut_fields_line_delim and is used only when the field delimiter equals
the record delimiter, because -d $'\n' and -z -d '' have different
semantics from normal line-based field splitting.
As part of that, I also folded the duplicated “start selected field”
logic into a shared helper, and renamed the byte-search helpers to match
their broader use. The current dispatcher in src/cut.c is now:
whitespace parser, then line- delimiter field mode, then byte-search
field mode, then the decoded multibyte parser.
* src/cut.c: Prefer strstr() over memmem() as the former is
optimized per platforma on GLIBC, where as the latter is only
optimized for S390 as of glibc-2.42
TODO: look at merging cut_fields and cut_fields_mb_utf8
as they're both byte search
TODO: See why a bit slower than old code
$ time src/cut.old -f1 -dç mb.in >/dev/null
real 0m0.136s
user 0m0.096s
sys 0m0.039s
$ time src/cut.new -f1 -dç mb.in >/dev/null
real 0m0.170s
user 0m0.139s
sys 0m0.030s
Support ignoring leading and trailing whitespace.
E.g. this matches awk's default field splitting mode.
* src/cut.c
* tests/cut/cut.pl: Add test cases.
To improve compatibility with toybox/busybox scripts.
* doc/coreutils.texi (cut invocation): Add -O description.
* src/cut.c: Support -O as well as --output-delimiter
* tests/cut/cut.pl: Adjust one case to use -O.
* src/cut.c: 160 fewer lines
Helpers extracted (replacing repeated inline patterns):
- write_line_delim(), write_pending_line_delim(), reset_item_line()
- line boundary code used by cut_bytes{,no_split}, cut_characters
- write_selected_item()
- output-delimiter + write logic used by all three byte/char functions
- reset_field_line()
- field line reset used by cut_fields_mb_any
Field functions unified via cut_fields_mb_any(stream, whitespace_mode):
- struct mbfield_parser encapsulates the whitespace vs.
fixed-delimiter state (saved char, mode flag)
- mbfield_get_char() - dispatches to saved-char or direct read
- mbfield_terminator()
- returns FIELD_{DATA,DELIMETER,LINE_DELIMITER} based on mode
- read_mb_field_to_buffer()
- replaces the two duplicated first-field buffering loops
- scan_mb_field(mbbuf, parser, pending, write_field)
- replaces the four duplicated field scan loops
(print+skip × two modes) with a single function and a write_field bool
- cut_fields_mb and cut_fields_ws are now trivial wrappers
Both the i18n patch and FreeBSD/macOS support this option.
They do differ in behavior somewhat as the i18n patch
may output more bytes than requested.
$ printf '\xc3\xa9b\n' | i18n-cut -n -b1
é
There is also a bug in the i18n patch with multi-byte
at the start of a line:
$ printf '\xc3\xa9b\n' | i18n-cut -n -b1-2
éb
We follow the FreeBSD behavior since it seems more
useful to have -b be a hard limit, rather than a soft limit.
This also reduces the possibility of duplicate character output
with separate cut invocations with non overlapping byte ranges.
* src/cut.c (cut_bytes_no_split): A new function
similar to cut_characters, to handle multi-byte characters
with byte limit semantics.
* tests/cut/cut.pl: Add test cases.
Note this is a slight divergence from the i18n patch
as that switched to uni-byte for any single byte delimiter
that is not valid multi-byte.
That results in possibly splitting in the middle of
a valid multi-byte character.
Instead we only split on a single byte when they're
not part of a multi-byte character.
* src/cut.c
Adjust all utils to:
- Have a separate translation per option
- Use either oputs() or oprintf(), instead of fputs() or printf().
- Use more standard alignment as discussed in previous commits.
- Various tweaks to descriptions.
Use new Gnulib streq function instead of rolling our own macro.
* bootstrap.conf (gnulib_modules): Add stringeq.
* src/rm.c (main): Don’t assume streq is a macro that expands to (...),
as it is now a function.
* src/system.h:
* tests/df/no-mtab-status.sh, tests/df/skip-duplicates.sh:
(STREQ): Remove. All uses replaced by streq.
Update to latest gnulib with new copyright year.
Run "make update-copyright" and then...
* gnulib: Update included in this commit as copyright years
are the only change from the previous gnulib commit.
* tests/init.sh: Sync with gnulib to pick up copyright year.
* bootstrap: Likewise.
* tests/sample-test: Adjust to use the single most recent year.
Update to latest gnulib with new copyright year.
Run "make update-copyright" and then...
* gnulib: Update included in this commit as copyright years
are the only change from the previous gnulib commit.
* tests/init.sh: Sync with gnulib to pick up copyright year.
* bootstrap: Manually update copyright year,
until we fully sync with gnulib at a later stage.
* tests/sample-test: Adjust to use the single most recent year.
* src/cut.c: Complete the error-handling improvements started in
commit e0a4a60af5, by adding a couple of remaining checks for putchar().
While there, sprinkle a few rather useful comments, and perform a few
small code cleanups, to make the code and the comments more uniform
and more conformant to the official coding style. Also make the help
message slightly more uniform.
* src/cut.c (cut_bytes): Diagnose errors from fwrite() and putchar().
(cut_fields): Likewise.
* tests/misc/write-errors.sh: Enable the test for cut,
and augment to cover both cut_bytes() and cut_fields().
* NEWS: Mention the improvement.
* src/cut.c (cut_file):
* src/nl.c (nl_file): Pacify GCC Bug#109613 in a better way, by
narrowing the coverage of the ‘assume’ so that bugs in the
no-longer-covered part are not masked.
Now that Gnulib’s ‘error’ module does proper static checking
for not returning, we need no longer use the ‘die’ macro.
This makes code easier to read for people that are used to ‘error’.
* cfg.mk (error_fns, exclude_file_name_regexp): Remove ‘die’.
(sc_die_EXIT_FAILURE): Remove.
* src/die.h: Remove. All includes removed. All calls to ‘die’
changed back to calls to ‘error’.
* src/install.c (get_ids): Use quoteaf (problem found with
make syntax-check).
* src/system.h: Include error.h, since some of our macros call ‘error’.
Stop including error.h elsewhere.
This modernizes the source code somewhat, to take advantage
of advances in GCC over the years, and Gnulib’s ‘assure’ module.
Include assure.h in files that now need it.
Do not include assert.h directly; it’s no longer needed.
* bootstrap.conf (gnulib_modules): Add ‘assure’.
* gl/lib/randread.c (randread_error):
* src/chmod.c (describe_change):
* src/chown-core.c (describe_change):
* src/cp.c (decode_preserve_arg):
* src/head.c (diagnose_copy_fd_failure):
* src/ls.c (parse_ls_color):
* src/od.c (decode_one_format):
* src/split.c (main):
* src/test.c (binary_operator, posixtest):
Prefer affirm to abort, since it has better diagnostics in the
normal case and better performance with -DNDEBUG.
* gl/lib/xdectoint.c, src/die.h: Include stddef.h, for unreachable.
* gl/lib/xdectoint.c: Do not include verify.h; no longer needed.
* gl/lib/xdectoint.c (__xnumtoint):
* src/die.h (die):
Prefer C23 unreachable () to assume (false).
* gl/lib/xfts.c (xfts_open):
* src/basenc.c (base32hex_encode):
* src/copy.c (abandon_move, copy_internal, valid_options):
* src/cut.c (cut_fields):
* src/df.c (alloc_field, decode_output_arg, get_dev):
* src/du.c (process_file, main):
* src/echo.c (usage):
* src/factor.c (udiv_qrnnd, mod2, gcd2_odd, factor_insert_large)
(mulredc2, factor_using_pollard_rho, isqrt2, div_smallq)
(factor_using_squfof):
* src/iopoll.c (iopoll_internal, fwrite_wait):
* src/join.c (add_field):
* src/ls.c (dev_ino_pop, main, gobble_file, sort_files):
* src/mv.c (do_move):
* src/od.c (decode_format_string, read_block, dump, main):
* src/remove.c (rm):
* src/rm.c (main):
* src/sort.c (stream_open):
* src/split.c (next_file_name, lines_chunk_split):
* src/stdbuf.c (main):
* src/stty.c (set_speed):
* src/tac-pipe.c (line_ptr_decrement, line_ptr_increment):
* src/touch.c (touch):
* src/tr.c (find_bracketed_repeat, get_next)
(validate_case_classes, get_spec_stats, string2_extend, main):
* src/tsort.c (search_item, tsort):
* src/wc.c (main):
Prefer affirm to assert, as it allows for better static
checking when compiling with -DNDEBUG.
* src/chown-core.c (change_file_owner):
* src/df.c (get_field_list):
* src/expr.c (printv, null, tostring, toarith, eval2):
* src/ls.c (time_type_to_statx, calc_req_mask, get_funky_string)
(print_long_format):
* src/numfmt.c (simple_strtod_fatal):
* src/od.c (decode_one_format):
* src/stty.c (mode_type_flag):
* src/tail.c (xlseek):
* src/tr.c (is_char_class_member, get_next, get_spec_stats)
(string2_extend):
Prefer unreachable () to abort () or assert (false) when merely
pacifying the compiler, e.g., in a switch statement on an enum
where all cases are covered.
* src/copy.c (valid_options): Now returns void; the bool was useless.
Caller no longer needs to assert.
* src/csplit.c (find_line):
* src/expand-common.c (next_file):
* src/shred.c (incname):
* src/sort.c (main):
* src/tr.c (append_normal_char, append_range, append_char_class)
(append_repeated_char, append_equiv_class):
* src/tsort.c (search_item):
Omit assert, since the hardware will check for us.
* src/df.c (header_mode): Now the enum type it should have been.
* src/du.c (process_file):
* src/ls.c (assert_matching_dev_ino):
* src/tail.c (valid_file_spec):
* src/tr.c (validate_case_classes):
Mark defns with MAYBE_UNUSED if they’re not used when -DNDEBUG.
* src/factor.c (prime_p, prime2_p, mp_prime_p): Now ATTRIBUTE_PURE.
Prefer affirm to error+abort. No need to translate this diagnostic.
* src/fmt.c (get_paragraph):
* src/stty.c (display_changed, display_all, sane_mode):
* src/who.c (idle_string):
Prefer assume to assert, since the goal is merely pacification
and assert doesn’t pacify anyway if -DNDEBUG is used.
* src/join.c (decode_field_spec):
Omit unreachable abort.
* src/ls.c (assert_matching_dev_ino, main):
* src/tr.c (get_next):
Prefer assure to assert, since the check is relatively expensive
and won’t help static analysis.
* src/ls.c (main):
Prefer static_assert to assert of a constant expression.
(format_inode): Redo to make it clear that buflen doesn’t matter,
and that buf must have a certain number of bytes. All callers changed.
This pacifies -Wformat-overflow.
* src/od.c (decode_one_format):
Omit an assert that tested for obviously undefined behavior,
as the compiler could optimize it away anyway.
* src/od.c (decode_one_format, decode_format_string):
Prefer ATTRIBUTE_NONNULL to runtime checking.
* src/stat.c: Do not include <stddef.h> since system.h does that now.
* src/sync.c (sync_arg):
Prefer unreachable () to assert (true), which was a typo.
* src/system.h: Include stddef.h, for unreachable.
* src/tail.c (xlseek): Simplify by relying on ‘error’ to exit.
Update to latest gnulib with new copyright year.
Run "make update-copyright" and then...
* tests/init.sh: Sync with gnulib to pick up copyright year.
* bootstrap: Manually update copyright year,
until we fully sync with gnulib at a later stage.
* tests/sample-test: Adjust to use the single most recent year.
* src/cut.c (enum operating_mode, operating_mode)
(output_delimiter_specified, cut_stream):
Remove; no longer needed.
(output_delimiter_default): New static var. Code can now
use ‘output_delimiter_string != output_delimiter_default’
instead of ‘output_delimiter_specified’.
(cut_file): New arg CUT_STREAM. Caller changed.
(main): Simplify. Coalesce duplicate code. Redo to avoid need
for IF_LINT, or for the static var. No need to xstrdup optarg.