One day I'll automatise it to happen with each auto indent. ..

FWIW I seem to remember that perltidy can't do this.

It aligns the assignment symbol = for me, too, which I don't want. Do you have any special align-rules-alist to prevent that?

#                     here,  instead of here
#                      v           V
my %periods            =  (Mercury => { orbital => 0.24,  rotation => 
+58.64 },
               Venus   => { orbital => 0.62,  rotation => -243.02 },
               Earth   => { orbital => 1.00,  rotation => 1.00 },
               Mars    => { orbital => 1.88,  rotation => 1.03 },
               Jupiter => { orbital => 11.86, rotation => 0.41 },
               Saturn  => { orbital => 29.46, rotation => 0.43 },
               Uranus  => { orbital => 84.01, rotation => -0.72 },
               Neptune => { orbital => 164.8, rotation => 0.67 },
              );
[download]

Moving the structure from the assignment line doesn't help:

my %periods            =  (
               Mercury => { orbital => 0.24,  rotation => 58.64 },
               Venus   => { orbital => 0.62,  rotation => -243.02 },
# ...
[download]

($q=q:Sq=~/;[c](.)(.)/;chr(-||-|5+lengthSq)`"S|oS2"`map{chr |+ord
}map{substrSq`S_+|`|}3E|-|`7**2-3:)=~y+S|`+$1,++print+eval$q,q,a,
[download]

I have the same problem, which I normally solve manually.

But this could be solved with separate rules for = , =~ and => and/or "sections separations".

please note how this works

my $h_periods =
  {
    Mercury => { orbital => 0.24,  rotation => 58.64 },
    Venus   => { orbital => 0.62,  rotation => -243.02 },
    Earth   => { orbital => 1.00,  rotation => 1.00 },
    Mars    => { orbital => 1.88,  rotation => 1.03 },
    Jupiter => { orbital => 11.86, rotation => 0.41 },
    Saturn  => { orbital => 29.46, rotation => 0.43 },
    Uranus  => { orbital => 84.01, rotation => -0.72 },
    Neptune => { orbital => 164.8, rotation => 0.67 },
};
[download]

That's because { is a section separator while ( is not

   `separate'  Each rule can define its own section separator, which
           describes how to identify the separation of "sections"
           within the region to be aligned.  Setting the `separate'
           attribute overrides the value of `align-region-separate'
           (see the documentation of that variable for possible
           values), and any separation argument passed to `align'.
[download]

using M-x customize-group RET align RET will facilitate customizing (it provides a textual UI for lisp data)

the rule in question in align-rules-list is

 Alignment rule:
            Title: perl-assignment
            Required attributes:
            Regexp:
            (Regular expression to match)
            Choice: Value Menu Regexp: [^=!^&*-+<>/|     
]\(\s-*\)=[~>]?\(\s-*\)\([^>=     
]\|$\)
            Optional attributes:
            INS DEL Choice: Value Menu Paren group:
                        (Parenthesis group to use)
                        Choice: Value Menu Repeat:
                        INS DEL Integer: 1
                        INS DEL Integer: 2
                        INS
            INS DEL Choice: Value Menu Modes:
                        (Modes where this rule applies)
                        Lisp expression: align-perl-modes
            INS DEL Choice: Value Menu To Tab Stop:
                        (Should rule align to tab stops)
                        Boolean: Toggle  off (nil)
            INS
[download]

this [~>]? part should be deleted or replaced.

(BTW: not sure what the -* is intended to do...)

Cheers Rolf
_{(addicted to the Perl Programming Language and ☆☆☆☆ :)
Je suis Charlie!}

The default alignment rules in emacs are meant to handle = assignment not => fat comma.

So task number one would be to separate these cases.

Secondly we need to define region separators.

as a demo: resetting the align-region-separate to handle parentheses helped

(snippet from customize: Align Region Separate: Value Menu Regexp defines section boundaries: [][{}])

to format

$a = {
    bla => 1,
    trax => [
        bla2 => {
            sauh1 => 1666,
            3 => 2,
        },
    ],
    #
    bla => 1,
    trallxa => 2,
};
[download]

to this, hence handling different levels accordingly

$a = {
    bla  => 1,
    trax => [
        bla2 => {
            sauh1 => 1666,
            3     => 2,
        },
    ],
    # new section
    bla     => 1,
    trallxa => 2,
};
[download]

but this doesn't help with this because the separators will be inside each line

my $h_periods = {
    Mercury => { orbital => 0.24, rotation => 58.64 },
    Venus => { orbal => 0.62, rotation => -243.02 },
    Earth => { orbital => 1.00, rotation => 1.00 },
    Mars => { orbital => 1.88, rotation => 1.03 },
    Jupiter => { orbital => 11.86, rotation => 0.41 },
    Saturn => { orbital => 29.46, rotation => 0.43 },
    Uranus => { orbital => 84.01, rotation => -0.72 },
    Neptune => { orbital => 164.8, rotation => 0.67 },
};
[download]

The only way to solve this is to use indentation to distinguish groups.

According to the docs it's possible to use a function call back instead of a regex, but documentation is sparse.

I wanted to document my findings before getting lost in other projects ... ;-)

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery FootballPerl is like chess, only without the dice}

> FWIW I seem to remember that perltidy can't do this.

wrong see http://perltidy.sourceforge.net/node005.html

I think I confused it with Data::Dump and Data::Dumper which do not align (fat) commas.

update

But I'm still not convinced about perltidy's performance for interactive use from within an IDE.

Cheers Rolf
_{(addicted to the Perl Programming Language and ☆☆☆☆ :)
Je suis Charlie!}

There is room for much improvement...

1. selecting the region:

a combination of backward-up-list and mark-sexp will select the surrounding structure (repeating will extend the selection till statement (ie. ';') boundary.

(please note that cperl-mode injects here some semantic knowledge about the code's structure which can't be reflected in a simple regexp rule)

2. the rule for align-current has the advantage to harmonize the surrounding whitespaces too. And align-regexp seems to allow the same extended logic.

So combining both steps bound to a key should make alignment of fat-commas quite easy.

I experimented with a kmacro and it seemed quite nice. :)