bug#39978: Weird substitution of square brackets

[Assaf Gordon]

tag 39978 notabug
close 39978
stop

Hello,

The commands you list below are correct and expected.
They follow POSIX's basic regular expression's "bracket expression"
behavior.

Note the following POSIX rules:
https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html#tag_09_03
    "9.3.3 - The <period>, <left-square-bracket>, and <backslash> shall be
    special except when used in a bracket expression"
and
    "9.3.5(1) - [...] The <right-square-bracket> ( ']' ) shall lose its
     special meaning and represent itself in a bracket expression if it
     occurs first in the list [...] Otherwise, it shall terminate the
     bracket expression"

On Sat, Mar 07, 2020 at 10:17:28PM +0200, Evangelos Tsagkas wrote:
> $ echo "object[0])" | sed 's/[\]\[]/./g'
> object[0])
> $ echo "object[0])" | sed 's/[\]]/./g'
> object[0])
> $ echo "object[0])" | sed 's/[\[]/./g'
> object.0])
> $ echo "object[])" | sed 's/[\]\[]/./g'
> object[])
> $ echo "object[])" | sed 's/[\[\]]/./g'
> object.)

To illustrate:

    # <backslash> in a bracket-expression is literal, not special
    $ echo '\' | sed 's/[\]/X/'

    # <left-square-bracket> inside a bracket-expression is literal, not special
    $ echo '[' | sed 's/[[]/X/'
    X

    # Combine the above two rules, and even though this might be
    # visually confusing, the bracket-expression only matches literal
    # <backslash> and <left-square-bracket>, regardless of how many
    # times they repeat in the bracket expression
    $ echo '\[' | sed 's/[\[\[\[[\]/X/g'
    XX

    # Here <backslash>-<left-square-bracket> causes the
    # <left-square-bracket> to become literal, not special:
    $ echo '[' | sed 's/\[/X/'

    # <right-square-bracket> appearing without a preceeding *special*
    # <left-square-bracket> is also literal:
    $ echo ']' | sed 's/]/X/'
    X

    # And thus:
    $ echo '[]' | sed 's/\[]/X/'
    X

I hope these explain how you got the above results.

As such, I'm closing this as "not a bug".
Discussion can continue by replying to this thread.

regards,
 - assaf