Linux Shell Help and Glob Patterns

  • Description: How to read built-in help on Linux (man, info, whatis, --help, apropos, tldr) and how shells expand glob patterns — wildcards, character classes, brace expansion — plus how globs differ from regular expressions.
  • My Notion Note ID: K2B-3-1
  • Created: 2020-06-03
  • Updated: 2026-05-19
  • License: Reuse is very welcome. Please credit Yu Zhang and link back to the original on yuzhang.io

Table of Contents

1. Reading the Help System

  • Four+ tiers, in order of increasing depth:
    • whatis <cmd> → one-line summary from the man database. Fastest sniff.
    • <cmd> --help / <cmd> -h → usage block printed by the program itself. Variable quality; some tools (ls, grep) have rich --help; minimal tools have none.
    • man <cmd> → full manual page. The default deep reference on Unix.
    • info <cmd> → GNU Texinfo manual — longer-form, hyperlinked, used by GNU coreutils/binutils. Often more detailed than man for those tools.
  • Auxiliary:
    • apropos <keyword> → search the whatis database by keyword. Same as man -k. Use when you can't remember the command name.
    • type <cmd> / which <cmd> → tells you whether a name is a builtin, alias, function, or external binary, and where the binary lives.
    • help <builtin> → shell builtin docs (Bash). man bash covers them too but is huge.
    • tldr <cmd> → community-maintained example-first cheat sheets (separate install). Skips prose, shows common invocations.

2. Conventions in Synopsis Lines

Same convention across man, --help, and POSIX docs:

Notation Meaning
[option] optional
<replace> or ARG placeholder — substitute a real value
| mutually exclusive choice
... one or more (repeatable)
[ARG]... zero or more
  • Example: ls [OPTION]... [FILE]... → both options and files are optional and may be repeated.

3. Navigating man

  • man <cmd> opens the manual in a pager (less by default).
  • Inside the pager:
    • /keyword — forward search. ?keyword — backward search.
    • n / N — next / previous match.
    • Space / b — page down / up. g / G — top / bottom.
    • q — quit.
  • Man sections (the number in ls(1), open(2)):
    • 1 user commands
    • 2 system calls (kernel API)
    • 3 library functions (libc, libm, …)
    • 5 file formats and configuration files (man 5 passwd)
    • 7 miscellaneous (protocols, glob, regex)
    • 8 sysadmin commands
  • Disambiguate when a name lives in multiple sections: man 2 open vs man 1 open.
  • man -k <regex> ≡ apropos (search short descriptions).
  • man -f <name> ≡ whatis (exact name match).

4. Navigating info

  • GNU's hyperlinked manual format. Pages are organized as a tree of nodes.
  • Key bindings:
    • PageUp / PageDown (or Space / Backspace) — scroll within a node.
    • Tab — jump to the next hyperlink.
    • Enter — follow the hyperlink under the cursor.
    • n / p — next / previous sibling node.
    • u — up one level. t — top of the manual.
    • s <keyword> — search across the whole manual.
    • l — go back (like a browser).
    • q — quit.
  • Many GNU tools have a one-screen man page and a real manual under info (e.g. info coreutils 'ls invocation').

5. Shell Globs vs Regex

Three pattern languages that look similar but behave differently:

Layer Where What it matches Examples
Shell glob (a.k.a. wildcard) shell expansion on filenames filename patterns, no regex semantics *.c, ??.txt, [abc]*
Regex (BRE / ERE) grep, sed, awk string patterns ^foo, a|b, (ab)+
PCRE grep -P, pcregrep, Perl, Python re richer regex (lookaround, lazy quantifiers, …) (?<=foo)bar, \d+?
  • Pitfall — * does not mean "any string" in glob the same way it does in regex. In glob, * matches any sequence of characters including the empty string, but only within a single path component (does not cross / unless globstar/** is enabled). In regex, * means "zero or more of the previous atom" — completely different.
  • The original [Y] note's example a*.b — under glob, that's "any filename starting with a and ending with .b" (e.g. a.b, abc.b, axyz.b). Under regex, it would mean "zero or more as followed by any single character followed by b" — almost never what you want for filenames.

6. Glob Syntax

Standard POSIX glob (supported by Bash, Zsh, Fish, etc.):

Pattern Matches
* any sequence of characters (within one path component)
? exactly one character
[abc] one character from the set
[a-z] one character in the range
[!abc] (or [^abc] in Bash/Zsh) one character not in the set

Bash extensions (shopt -s extglob):

Pattern Matches
?(p) zero or one occurrence of p
*(p) zero or more
+(p) one or more
@(p1|p2) exactly one of the alternatives
!(p) anything except p

Brace expansion (a separate mechanism — happens before glob expansion, even when no file matches):

  • cp file.{txt,bak} → cp file.txt file.bak
  • mkdir -p project/{src,test,docs} → three subdirs
  • echo {1..5} → 1 2 3 4 5. Step: {1..10..2} → 1 3 5 7 9.
  • echo {a..c}{1..2} → a1 a2 b1 b2 c1 c2.

Recursive glob (**) — Bash needs shopt -s globstar, Zsh is on by default:

  • **/*.py → every .py file anywhere under the current dir.

Hidden files:

  • * does not match leading-dot files (.bashrc, .git) by default. Use .* to include them, or shopt -s dotglob to make * match them.

7. References