[ruby-core:31484] [Bug #3616] IRB + readline incorrectly counts non-printing characters in prompt

Bug #3616: IRB + readline incorrectly counts non-printing characters in prompt
http://redmine.ruby-lang.org/issues/show/3616

Author: Sung Pae
Status: Open, Priority: Normal
ruby -v: ruby 1.9.2dev (2010-07-24 revision 28749) [x86_64-darwin10.4.0] 

When setting a prompt for IRB that contains terminal control characters:

    IRB.conf[:PROMPT][:COLOR] = {
      :PROMPT_I => "\e[0;1;32m>>> \e[0m",
      ...
    }
    IRB.conf[:PROMPT_MODE] = :COLOR

IRB includes "\e[...m" in the prompt length calculation, causing readline
functions like beginning-of-line to move to the incorrect place on the line:

    >>> 012345678901234567890
                             ^ caret
    <Press \C-a>
    >>> 012345678901234567890
                     ^ caret

Fortunately, GNU Readline offers a couple of constants to mark that a string
should be non-printing:

    # readline.h

    /* Definitions available for use by readline clients. */
    #define RL_PROMPT_START_IGNORE	'\001'
    #define RL_PROMPT_END_IGNORE	'\002'

Bash handles this issue by mapping `\[' and `\]' to RL_PROMPT_START_IGNORE and
RL_PROMPT_END_IGNORE inside of `PS*' prompt variables. It would be nice to offer
the same feature in IRB, or simply to surround all terminal escape sequences in
the IRB prompt with RL_PROMPT_*_IGNORE.

Note that this bug does not occur with libedit.


----------------------------------------
http://redmine.ruby-lang.org