Quickstart

% git clone --recurse-submodules https://github.com/drewr/zigutils
% ./bootstrap.sh
% zig build run

zsh-nix-env

Prints a compact label describing the active Nix shells, meant for a zsh precmd:

nix_info=$(zsh-nix-env)
[[ -n $nix_info ]] && nix_prompt="[${nix_info}]"

Active shells	Output
none	(empty)
nix-shell -p jq ripgrep	λjq:ripgrep
nix develop	❄
nix-shell inside nix develop	❄λjq:ripgrep

Each environment is a ShellScenario variant with a ShellDetector that inspects the environment. ActiveShells runs every detector, so more than one scenario can be active at once. To support a new environment, add a variant and its render arm to ShellScenario and a detector to shell_detectors in src/zsh-nix-env.zig.

zsh-hist

Delete entries matching a pattern from zsh history. It is a dry run by default; pass --delete to rewrite the file. A backup is written first and the original file permissions are preserved.

zsh-hist [options] PATTERN

Option	Meaning
-d, --delete	actually delete matching entries
-F, --fixed-strings	treat PATTERN as a literal string
-i, --ignore-case	case-insensitive match
-x, --line-regexp	PATTERN must match the whole command
-f, --file FILE	history file (default: $HISTFILE or ~/.zsh_history)
--no-backup	don't write a backup before deleting
-h, --help	show help

Examples:

% zsh-hist 'git status'            # preview matching entries
% zsh-hist -x 'git status'         # commands that are exactly "git status"
% zsh-hist -F 'npm run build'      # literal, not a regex
% zsh-hist -d '^cd /tmp'           # delete entries starting with "cd /tmp"

Plain and EXTENDED_HISTORY (: <ts>:<dur>;cmd) entries are handled, as are multi-line commands and zsh's metafied encoding of non-ASCII bytes. With --delete, the tool takes zsh's HISTFILE.LOCK first so a shell appending to the history during the rewrite cannot lose lines.

Shared-history caveat: open shells keep their own copy of the history in memory and may write it back later. After deleting entries, run fc -R in each open shell (or restart them) to reload the cleaned file.

regex (library)

src/regex.zig is a small pure-Zig regular-expression library used by zsh-hist. It has no IO and no dependencies beyond std. src/regex_test.zig covers it with example-based, generative and fuzz tests.

The public API is:

const regex = @import("regex");

var re = try regex.Regex.compile(gpa, "a+", .{ .case_insensitive = true });
defer re.deinit(gpa);
if (re.isMatch("BAA")) { /* ... */ }

Options also has literal (quote the whole pattern) and whole (anchor both ends). regex.escape(gpa, bytes) returns a pattern that matches bytes exactly. The parser's Node and parse() are exported for tests but are internal details.

Supported syntax:

• literals, . (any byte except newline), ^, $, \A, \Z
• alternation, (...) and (?:...)
• * + ? {m} {m,} {,n} {m,n}, with an optional lazy ? (it is accepted and ignored, since matching only answers yes or no)
• bracket classes [...] and [^...], with ranges and escapes
• \d \D \w \W \s \S \b \B \n \t \r \f \v, and \ before any punctuation

Patterns match bytes, so UTF-8 text works as its byte sequence. Matching uses a Thompson NFA, so it is linear in the input size for a fixed pattern and does not backtrack.

To use the library from another tool, add an import in build.zig:

const regex_mod = b.createModule(.{
    .root_source_file = b.path("src/regex.zig"),
    .target = target,
    .optimize = optimize,
});

const my_tool = b.addExecutable(.{
    .name = "my-tool",
    .root_module = b.createModule(.{
        .root_source_file = b.path("src/my-tool.zig"),
        .target = target,
        .optimize = optimize,
        .imports = &.{.{ .name = "regex", .module = regex_mod }},
    }),
});