feat(cli): add zxcgrep utility for searching compressed files

README.md

@@ -192,6 +192,9 @@ zxccat logs-1.zxc logs-2.zxc | grep ERROR
 # Read a compressed file page by page in less (POSIX installs)
 zxcless app.log.zxc
 
+# Search compressed files with grep's options (POSIX installs)
+zxcgrep -n ERROR logs-*.zxc
+
 # Benchmark mode (testing speed on your machine)
 zxc -b assets.tar
 ```

cmake/zxcInstall.cmake

@@ -82,9 +82,9 @@ if(ZXC_INSTALL)
                         zxc \"\$ENV{DESTDIR}${ZXC_SYMLINK_BINDIR}/${alias}\")
                 ")
             endforeach()
-            # "zxcless": less(1) on .zxc files, a shell script.
-            install(PROGRAMS src/cli/zxcless DESTINATION ${CMAKE_INSTALL_BINDIR})
-            foreach(alias unzxc zxccat zxcless)
+            # "zxcless" and "zxcgrep": less(1) and grep(1) on .zxc files, shell scripts.
+            install(PROGRAMS src/cli/zxcless src/cli/zxcgrep DESTINATION ${CMAKE_INSTALL_BINDIR})
+            foreach(alias unzxc zxccat zxcless zxcgrep)
                 install(CODE "
                     execute_process(COMMAND \"${CMAKE_COMMAND}\" -E create_symlink
                         zxc.1 \"\$ENV{DESTDIR}${ZXC_SYMLINK_MANDIR}/man1/${alias}.1\")

meson.build

@@ -254,12 +254,12 @@ if not meson.is_subproject() and get_option('build_cli')
         install_dir : get_option('bindir'),
       )
     endforeach
-    # "zxcless": less(1) on .zxc files, a shell script.
-    install_data('src/cli/zxcless',
+    # "zxcless" and "zxcgrep": less(1) and grep(1) on .zxc files, shell scripts.
+    install_data('src/cli/zxcless', 'src/cli/zxcgrep',
       install_dir : get_option('bindir'),
       install_mode : 'rwxr-xr-x',
     )
-    foreach alias : ['unzxc', 'zxccat', 'zxcless']
+    foreach alias : ['unzxc', 'zxccat', 'zxcless', 'zxcgrep']
       install_symlink(alias + '.1',
         pointing_to : 'zxc.1',
         install_dir : get_option('mandir') / 'man1',

src/cli/main.c

@@ -16,6 +16,7 @@
 
 #include <errno.h>
 #include <limits.h>
+#include <signal.h>
 #include <stdarg.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -1449,6 +1450,11 @@ int main(int argc, char** argv) {
     int to_stdout = 0;
     int cat_mode = 0;
 
+#ifndef _WIN32
+    // Reader gone (zxc -dc f | head): die quietly, even if the parent ignored SIGPIPE
+    signal(SIGPIPE, SIG_DFL);
+#endif
+
     /* As "unzxc", default to decompression; as "zxccat", to zxc -dcf on every
      * argument. An explicit mode flag below still wins. */
     {

src/cli/zxc.1

@@ -2,8 +2,8 @@
 .\"
 .TH "zxc" "1" "" "zxc" "User Commands"
 .SH NAME
-\f[B]zxc\f[R], \f[B]unzxc\f[R], \f[B]zxccat\f[R], \f[B]zxcless\f[R] \-
-Compress or decompress .zxc files
+\f[B]zxc\f[R], \f[B]unzxc\f[R], \f[B]zxccat\f[R], \f[B]zxcless\f[R],
+\f[B]zxcgrep\f[R] \- Compress or decompress .zxc files
 .SH SYNOPSIS
 \f[B]zxc\f[R] [\f[I]OPTIONS\f[R]] [\f[I]INPUT\-FILE\f[R]]
 [\f[I]OUTPUT\-FILE\f[R]]
@@ -15,15 +15,33 @@ Compress or decompress .zxc files
 input that does not start like a ZXC archive (see \f[B]\-f\f[R]) comes
 out unchanged.
 .PP
-\f[B]zxcless\f[R] [\f[I]LESS\-OPTIONS\f[R]] [\f[I]FILE\f[R]...]
+\f[B]zxcless\f[R] [\f[B]\-\-dict=\f[R]\f[I]FILE\f[R]]
+[\f[I]LESS\-OPTIONS\f[R]] [\f[I]FILE\f[R]...]
 shows files, or standard input, in \f[B]less\f[R](1), decompressing ZXC
 archives on the fly; other input displays unchanged.
+\f[B]\-\-dict\f[R] supplies the dictionary of archives that need one
+(see \f[B]\-D\f[R]).
 The \f[B]ZXC\f[R] environment variable names the \f[B]zxc\f[R] binary to
 use.
 .PP
+\f[B]zxcgrep\f[R] [\f[B]\-\-dict=\f[R]\f[I]FILE\f[R]]
+[\f[I]GREP\-OPTIONS\f[R]] [\f[B]\-e\f[R]] \f[I]PATTERN\f[R]
+[\f[I]FILE\f[R]...]
+runs \f[B]grep\f[R](1) on files, or standard input, decompressing ZXC
+archives on the fly; other input is searched unchanged.
+\f[B]\-\-dict\f[R] supplies the dictionary of archives that need one
+(see \f[B]\-D\f[R]).
+Options go before the pattern; those that select files (\f[B]\-r\f[R],
+\f[B]\-d\f[R], \f[B]\-D\f[R], \f[B]\-\-include\f[R],
+\f[B]\-\-exclude\f[R]) and \f[B]\-f \-\f[R] are not supported.
+The exit status is grep\(aqs: 0 if a line matched, 1 if none did, 2 on
+any error, including a damaged archive.
+\f[B]ZXC\f[R] and \f[B]GREP\f[R] name the binaries to use; grep must
+support \f[B]\-\-label\f[R] (GNU and BSD grep do).
+.PP
 \f[B]unzxc\f[R] and \f[B]zxccat\f[R] are symlinks to \f[B]zxc\f[R], and
-\f[B]zxcless\f[R] is a shell script; all three are installed on POSIX
-systems only.
+\f[B]zxcless\f[R] and \f[B]zxcgrep\f[R] are shell scripts; all four are
+installed on POSIX systems only.
 .SH DESCRIPTION
 \f[B]zxc\f[R] is a command\-line interface for the ZXC compression
 library, a high\-performance lossless compression algorithm optimized
@@ -284,6 +302,9 @@ zxccat part1.zxc part2.zxc | grep ERROR
 .PP
 \f[B]Read a compressed log page by page:\f[R] zxcless app.log.zxc
 .PP
+\f[B]Search compressed logs, with file names and line numbers:\f[R]
+zxcgrep \-n ERROR logs\-*.zxc
+.PP
 \f[B]Compress multiple files independently:\f[R] zxc \-m file1.txt
 file2.txt file3.txt
 .PP

src/cli/zxc.1.md

@@ -1,7 +1,7 @@
 # zxc(1)
 
 ## NAME
-**zxc**, **unzxc**, **zxccat**, **zxcless** - Compress or decompress .zxc files
+**zxc**, **unzxc**, **zxccat**, **zxcless**, **zxcgrep** - Compress or decompress .zxc files
 
 ## SYNOPSIS
 **zxc** [*OPTIONS*] [*INPUT-FILE*] [*OUTPUT-FILE*]
@@ -10,9 +10,11 @@
 
 **zxccat** is equivalent to **zxc -dcf** applied to every *INPUT-FILE* in turn, writing them all to standard output; input that does not start like a ZXC archive (see **-f**) comes out unchanged.
 
-**zxcless** [*LESS-OPTIONS*] [*FILE*...] shows files, or standard input, in **less**(1), decompressing ZXC archives on the fly; other input displays unchanged. The **ZXC** environment variable names the **zxc** binary to use.
+**zxcless** [**--dict=***FILE*] [*LESS-OPTIONS*] [*FILE*...] shows files, or standard input, in **less**(1), decompressing ZXC archives on the fly; other input displays unchanged. **--dict** supplies the dictionary of archives that need one (see **-D**). The **ZXC** environment variable names the **zxc** binary to use.
 
-**unzxc** and **zxccat** are symlinks to **zxc**, and **zxcless** is a shell script; all three are installed on POSIX systems only.
+**zxcgrep** [**--dict=***FILE*] [*GREP-OPTIONS*] [**-e**] *PATTERN* [*FILE*...] runs **grep**(1) on files, or standard input, decompressing ZXC archives on the fly; other input is searched unchanged. **--dict** supplies the dictionary of archives that need one (see **-D**). Options go before the pattern; those that select files (**-r**, **-d**, **-D**, **--include**, **--exclude**) and **-f -** are not supported. The exit status is grep's: 0 if a line matched, 1 if none did, 2 on any error, including a damaged archive. **ZXC** and **GREP** name the binaries to use; grep must support **--label** (GNU and BSD grep do).
+
+**unzxc** and **zxccat** are symlinks to **zxc**, and **zxcless** and **zxcgrep** are shell scripts; all four are installed on POSIX systems only.
 
 ## DESCRIPTION
 **zxc** is a command-line interface for the ZXC compression library, a high-performance lossless compression algorithm optimized for maximum decompression throughput.
@@ -143,6 +145,9 @@ Archives can be concatenated (`cat a.zxc b.zxc > ab.zxc`): **-d** decodes them o
 **Read a compressed log page by page:**
   zxcless app.log.zxc
 
+**Search compressed logs, with file names and line numbers:**
+  zxcgrep -n ERROR logs-*.zxc
+
 **Compress multiple files independently:**
   zxc -m file1.txt file2.txt file3.txt
 

src/cli/zxcgrep

@@ -0,0 +1,133 @@
+#!/bin/sh
+# SPDX-License-Identifier: BSD-3-Clause
+# ZXC - High-performance lossless compression
+#
+# Copyright (c) Bertrand Lebonnois and contributors.
+#
+# zxcgrep: grep(1) on .zxc files, decompressed on the fly, other files as is.
+# grep's options go before the pattern; those that pick files are refused.
+# --dict=FILE names a zxc dictionary.
+
+zxc=${ZXC:-zxc}
+grep=${GREP:-grep}
+
+usage() {
+    echo "Usage: zxcgrep [GREP-OPTIONS] [-e] PATTERN [FILE...]"
+}
+
+die() {
+    echo "zxcgrep: $1" >&2
+    exit 2
+}
+
+# qv = $1 quoted for eval
+q() {
+    qs=$1 qv=''
+    while :; do
+        case $qs in
+        *\'*) qv="$qv${qs%%\'*}'\\''"; qs=${qs#*\'} ;;
+        *) qv="'$qv$qs'"; return ;;
+        esac
+    done
+}
+
+add() {
+    q "$1"
+    opts="$opts $qv"
+}
+
+# grep's stdin is the data
+nostdin='-f -: patterns from standard input not supported'
+
+opts='' pat='' no_name='' quiet='' silent='' dict=''
+while [ "$#" -gt 0 ]; do
+    case $1 in
+    --) shift; break ;;
+    --help) usage; exit 0 ;;
+    --recursive | --dereference-recursive | --directories* | --devices* | \
+    --include* | --exclude*) die "${1%%=*}: not supported" ;;
+    --file=-) die "$nostdin" ;;
+    --dict=*) dict=${1#--dict=}; shift; continue ;;
+    --dict)
+        [ "$#" -ge 2 ] || die "$1: missing argument"
+        dict=$2; shift 2; continue ;;
+    --regexp=* | --file=*) pat=1 ;;
+    --regexp | --file)
+        pat=1
+        [ "$#" -ge 2 ] || die "$1: missing argument"
+        [ "$1 $2" != "--file -" ] || die "$nostdin"
+        add "$1"; shift ;;
+    --after-context | --before-context | --context | --max-count | --label | \
+    --binary-files | --group-separator)
+        [ "$#" -ge 2 ] || die "$1: missing argument"
+        add "$1"; shift ;;
+    --quiet | --silent) quiet=1 ;;
+    --no-messages) silent=1 ;;
+    --no-filename) no_name=1 ;;
+    --*) ;;
+    -?*)
+        # -inA 3, -ie PAT: a letter taking a value ends the cluster
+        o=${1#-}
+        while [ -n "$o" ]; do
+            c=${o%"${o#?}"}; o=${o#?}
+            case $c in
+            [rRdD]) die "-$c: not supported" ;;
+            [ef]) pat=1 ;;
+            h) no_name=1 ;;
+            q) quiet=1 ;;
+            s) silent=1 ;;
+            esac
+            case $c in
+            [ABCefm])
+                if [ -z "$o" ]; then
+                    [ "$#" -ge 2 ] || die "-$c: missing argument"
+                    o=$2; add "$1"; shift
+                fi
+                [ "$c$o" != f- ] || die "$nostdin"
+                break ;;
+            esac
+        done ;;
+    *) break ;;
+    esac
+    add "$1"; shift
+done
+
+if [ -z "$pat" ]; then
+    [ "$#" -gt 0 ] || { usage >&2; exit 2; }
+    add -e; add "$1"; shift
+fi
+[ "$#" -gt 0 ] || set -- -
+[ "$#" -gt 1 ] && [ -z "$no_name" ] && opts="-H$opts"
+
+sigpipe() { [ "$1" -gt 128 ] && [ "$(kill -l "$1" 2>/dev/null)" = PIPE ]; }
+
+unzxc() {
+    if [ -n "$dict" ]; then "$zxc" -dcf -D "$dict" -- "$1"; else "$zxc" -dcf -- "$1"; fi
+}
+
+# Status as grep: 0 match, 1 none, 2 error. -s silences zxc too.
+if [ -n "$silent" ]; then exec 5>/dev/null; else exec 5>&2; fi
+exec 3>&1
+st=1
+for f do
+    label=''
+    if [ "$f" != - ]; then q "--label=$f"; label=$qv; fi
+    # Status: grep's from the pipeline, zxc's on fd 4
+    z=$(
+        exec 4>&1
+        eval "set -- $label $opts -"
+        { unzxc "$f" 2>&5 4>&-; echo "$?" >&4; } | "$grep" "$@" 4>&- >&3
+    )
+    g=$?
+    # grep's errors (bad pattern, output closed) hold for every file
+    [ "$g" -gt 1 ] && exit "$g"
+    # -q: the first match settles it
+    [ "$g" -eq 0 ] && [ -n "$quiet" ] && exit 0
+    # SIGPIPE on zxc: grep stopped reading (-l, -m...)
+    if [ "$z" -ne 0 ] && ! sigpipe "$z"; then
+        st=2
+    elif [ "$g" -eq 0 ] && [ "$st" -eq 1 ]; then
+        st=0
+    fi
+done
+exit "$st"

src/cli/zxcless

@@ -5,18 +5,35 @@
 # Copyright (c) Bertrand Lebonnois and contributors.
 #
 # zxcless: less(1) on .zxc files, decompressed on the fly by this script rerun as
-# the LESSOPEN filter. Needs less 451+.
+# the LESSOPEN filter. Needs less 451+. --dict=FILE names a zxc dictionary.
 
 zxc=${ZXC:-zxc}
 
 if [ "$#" -eq 2 ] && [ "$1" = --zxcless-filter ]; then
+    f=$2
+    set -- -dcq
+    [ -z "$ZXCLESS_DICT" ] || set -- "$@" -D "$ZXCLESS_DICT"
     # -f for stdin only: a named file zxc can't decode, less opens itself.
-    if [ "$2" = - ]; then
-        exec "$zxc" -dcfq -
+    if [ "$f" = - ]; then
+        exec "$zxc" -f "$@" -
     fi
-    exec "$zxc" -dcq -- "$2"
+    exec "$zxc" "$@" -- "$f"
 fi
 
+# --dict is ours, the rest is less's: the filter reads it from the environment
+want=''
+for a do
+    shift
+    if [ -n "$want" ]; then ZXCLESS_DICT=$a want=''; continue; fi
+    case $a in
+    --dict=*) ZXCLESS_DICT=${a#--dict=}; continue ;;
+    --dict) want=1; continue ;;
+    esac
+    set -- "$@" "$a"
+done
+[ -z "$want" ] || { echo "zxcless: --dict: missing argument" >&2; exit 1; }
+export ZXCLESS_DICT
+
 # "||": empty output with status 0 (an empty archive) shows empty; "-": stdin too.
 export LESSOPEN="||-sh \"$0\" --zxcless-filter %s"
 unset LESSCLOSE