daniil-berg
7166d9196a
The macro form `"\e[A": "\C-x\C-p\C-e"` ran history-search-backward then end-of-line to mimic zsh's cursor-at-end. But history-search-* takes its prefix from start-of-line to point, and `\C-e` moves point to the end, so the next Up searched for the whole recalled line, found nothing, and left Up/Down dead after one press. Bind arrows straight to the search functions; pure readline has no history-search-end widget, so cursor lands at the prefix boundary. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
21 lines
1021 B
Plaintext
21 lines
1021 B
Plaintext
# Readline config (bash line editing + any readline program: python, psql...).
|
|
# Set via $INPUTRC; see shell/exports. zsh ignores this (uses ZLE); its
|
|
# equivalent prefix-search bindings live in zsh/.zshrc.
|
|
#
|
|
# Pull in distro defaults; a custom INPUTRC otherwise replaces them entirely.
|
|
$include /etc/inputrc
|
|
|
|
# zsh-style prefix history search: type a prefix, Up/Down walk matching
|
|
# history entries. Empty prefix = previous/next.
|
|
#
|
|
# Bind arrows straight to the search functions. Do NOT wrap in a macro that
|
|
# appends end-of-line (\C-e) to mimic zsh's cursor-at-end: \C-e moves point to
|
|
# the line end, and history-search-* takes its prefix from start-of-line to
|
|
# point -- so the next press searches for the whole recalled line, finds
|
|
# nothing, and Up/Down go dead. Pure readline has no history-search-end widget;
|
|
# cursor lands at the prefix boundary, which is readline's intended behavior.
|
|
"\e[A": history-search-backward
|
|
"\e[B": history-search-forward
|
|
"\eOA": history-search-backward
|
|
"\eOB": history-search-forward
|