Trimming indentation

dbohdan 2014-06-14: I wanted to have a multiline string help message indented to the level of the rest of the code. The following procedure unindents it appropriately but keeps the relative indentation.

# Obsolete code removed. See the current version below.

Use example

[...]
    if {[llength $args] == 0} {
        # Print help
        puts -nonewline [
            trim-indentation {
                Command pipelines for interactive programming.

                usage:
                    |> cmd1 |> cmd2 _ |> cmd3 #0 ?-debug?
                or
                    |> { cmd1 |> cmd2 $_ |> cmd3 $pipe(0) } ?-debug?

                See https://wiki.tcl-lang.org/17419 for more.
            }
            return
        ]
        # End help
    }
[...]

Output

Command pipelines for interactive programming.

usage:
    |> cmd1 |> cmd2 _ |> cmd3 #0 ?-debug?
or
    |> { cmd1 |> cmd2 $_ |> cmd3 $pipe(0) } ?-debug?

See https://wiki.tcl-lang.org/17419 for more.

PYK 2014-06-14: What does shift do? This implementation uses the simplistic method of counting the number of characters trimmed from the first line, and fails for cases where the first line is more indented than other lines. To do the job right, see textutil::undent. Also, as of 8.6, lmap is the built-in replacement for struct::list mapfor.

dbohdan 2014-06-14: The shift (an eltclsh built-in) is there because when the command is used like in the example above and the argument is split at newlines the first line becomes {}.

I wasn't aware of textutil::undent; thanks for suggesting it!

Out of curiosity I implemented trimming by the longest possible unindent in a functional style. Its advantage over textutil::undent is that it supports custom and multiple delimiters, although when multiple delimiters are specified it treats them as interchangeable. shifting is replaced with a proc called ltrim that removes empty items on both ends of a list. It is called to get rid of empty lines and lines of just the base level of indentation preceding and following the text, which I found useful for inline documents like the help message example above.

# Trim indentation in multiline quoted text.
proc trim-indentation {msg {whitespaceChars " "}} {
    set msgLines [split $msg "\n"]
    set maxLength [string length $msg]

    set regExp [subst -nocommands {([$whitespaceChars]*)[^$whitespaceChars]}]

    set indent [
        tcl::mathfunc::min {*}[
            struct::list mapfor x $msgLines {
                if {[regexp $regExp $x match whitespace]} {
                    string length $whitespace
                } else {
                    lindex $maxLength
                }
            }
        ]
    ]

    return [
        join [
            ltrim [
                struct::list mapfor x $msgLines {string range $x $indent end}
            ]
        ] "\n"
    ]
}

# Remove empty items at the beginning and the end of a list.
proc ltrim {list} {
    set first [lsearch -not -exact $list {}]
    set last [lsearch -not -exact [lreverse $list] {}]
    return [
        if {$first == -1} {
            list
        } else {
            lrange $list $first end-$last
        }
    ]
}

(I'm not using lmap here because I'm on 8.5.)

eltclsh > trim-indentation  "\n\n\n  a\n\n        b\n\n\n"
a

      b
eltclsh > trim-indentation  "\n\n\n  a\n\n        b\n  \n\n\n"
a

      b
eltclsh > trim-indentation  "\n\n\n  a\n\n        b\n   \n\n\n"
a

      b
 
eltclsh > trim-indentation  "-----a\n---b\n-----c\n" -
--a
b
--c