#csplit

Version 0.2.1 behavior is documented here. The support record separates local verification from published MoonX behavior; moonx examples use registry releases.

Split a file into sections at line numbers or basic regular expressions:

moonx cli/csplit input.txt 10 20
moonx cli/csplit -s -f section- input.txt '/^Chapter/' '{*}'

NUMBER splits before that absolute line; /REGEXP/ splits before a matching line and %REGEXP% discards the preceding section. Regex delimiters may be followed by a positive or negative line offset. {N} repeats the preceding pattern N additional times. Regex {*} repeats through the remaining input; numeric repetitions that reach beyond the last line report an error. The first operand - reads stdin. All input bytes, including NUL and an unterminated final line, are preserved in selected fragments.

Files are named xx00, xx01, … . -f/--prefix changes the prefix and -n/--digits sets the minimum number of suffix digits. Byte counts are printed for created files unless -s/--silent or -q/--quiet is selected. -z/--elide-empty-files omits empty fragments. Split failures remove the files created by that invocation unless -k/--keep-files is used.

-b/--suffix-format replaces the default digits with exactly one d/i/u/o/x/X integer conversion, literal text and %%. Width, precision, left/zero padding and applicable alternate/thousands flags follow the GNU C profile. Repeated formats use the last; -b determines layout independently of -n ordering. Invalid formats fail before file creation. Windows ordinary filesystem paths trim trailing spaces and dots. A generated name with either ending is therefore refused with status 2 before reading input; it never silently creates a differently named fragment. Verbatim-name support requires verified Native/Wasm creation, reading and observation of those exact names.

--suppress-matched omits the line at the offset-adjusted split boundary for regex, ignored regex and numeric patterns. Empty elision reuses the next output suffix; suppression preserves the separate regex-search cursor.

Input is consumed once into an owned temporary snapshot before creating fragments. The snapshot and its 64-bit line-offset index stay on disk; each index/data cache is at most 64 KiB. Numeric splits copy byte ranges in 64 KiB blocks. Regex searches retain only the current line, so their extra memory includes the largest matched or searched line, rather than every input record. Temporary disk use is the input byte count plus eight bytes per line boundary, including the initial boundary. Created output paths remain in memory for failure cleanup. Regex line storage and numeric pattern/offset parsing keep their existing signed 32-bit limits; file and index positions use checked signed 64-bit arithmetic.

The temporary directory and its files are exclusive to the invocation; snapshot files use mode 0600. They are removed after success, ordinary errors and cancellation observed by the runtime. A killed process or machine failure cannot guarantee that cleanup. Temp creation/write or input-read errors occur before fragment creation. After splitting begins, -k keeps completed or partial fragments; without it, the invocation removes only the fragment paths it created.

The shared C-byte BRE matcher compiles each expression once and supports capture backreferences, byte classes and word assertions. Repeated nullable captures follow the tested GNU candidate and capture-register selection, including counted repeats and references. See the workspace support record. Existing output paths are refused. Locale-sensitive matching is outside the selected profile. --help and --version provide command information.

#Standalone metadata

Standalone --version prints exactly csplit 0.2.1 followed by one newline. Standalone --help starts with Usage: csplit, lists both metadata options, and ends with a newline. Both write to stdout, leave stdin unread, and exit with status 0. Exact-version publication and verification remain recorded separately in the support record.