[Caml-list] Tuareg mode: Indentation of application arguments

[Caml-list] Tuareg mode: Indentation of application arguments

[Romain Bardou]

Hello Caml-list,

I have finally took the time to configure tuareg to correctly indent 
after a "| ... ->" clause. After several years of doing it by hand, this 
feels soooo much better :) To do this I had to add "(setq 
tuareg-use-smie t)" to my .emacs (which I found a little confusing, by 
the way).

However, now it's indentation of application arguments which is messed 
up (to my tastes). Let's say I am writing:

a_very_long_function_name argument_1 argument_2 argument_3 argument_4

Now I want to put argument_4 on a new line. Before, I would obtain:

a_very_long_function_name argument_1 argument_2 argument_3
   argument_4

which I liked. Now, I obtain:

a_very_long_function_name argument_1 argument_2 argument_3
                           argument_4

Note that if I put all arguments on their own lines, indentation does 
not depend on the function name:

a_very_long_function_name
   argument_1
   argument_2
   argument_3
   argument_4

I did not find any variable controlling this behavior in the tuareg 
group. Do you know how I can obtain my beloved behavior back? Or is it a 
bug, in which case I should submit a bug report?

The reason I don't like this behavior is that it causes several issues:
- changing one function name would imply that one would have to reindent 
everything (non-local indentation: bad!);
- the indentation is not simply "indentation of previous line + N" where 
N is the same everywhere (complex indentation rule: bad!);
- the indentation size may be very long (sudden-death indentation: bad!).

My version of tuareg is 2.0.6.

Cheers,

-- 
Romain Bardou

[Caml-list] Re: Tuareg mode: Indentation of application arguments

[Stefan Monnier]

>> Now I want to put argument_4 on a new line. Before, I would obtain:
>>
>> a_very_long_function_name argument_1 argument_2 argument_3
>>   argument_4
>>
>> which I liked. Now, I obtain:
>>
>> a_very_long_function_name argument_1 argument_2 argument_3
>>                           argument_4
>>
>> Note that if I put all arguments on their own lines, indentation does
>> not depend on the function name:
>>
>> a_very_long_function_name
>>   argument_1
>>   argument_2
>>   argument_3
>>   argument_4
>>
>> I did not find any variable controlling this behavior in the tuareg
>> group.

The nice thing about this behavior is that you don't need any config var
to control it: just split your lines at a different place instead.


        Stefan

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Romain Bardou]

> a configuration variable to change this behavior.  (BTW, do you know
> that M-q re-indent the current expression, so re-indenting after a name
> change is effortless.)

I know about M-q but I disagree that it is effortless: if you change a 
variable name, and this name is used in N places, you have to go to all 
those N places and re-indent there. Note that those N places can be 
different files.

Cheers,

-- 
Romain

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Kristopher Micinski]

On Fri, Sep 21, 2012 at 9:42 AM, Romain Bardou <romain.bardou@inria.fr> wrote:
>> a configuration variable to change this behavior.  (BTW, do you know
>> that M-q re-indent the current expression, so re-indenting after a name
>> change is effortless.)
>
>
> I know about M-q but I disagree that it is effortless: if you change a
> variable name, and this name is used in N places, you have to go to all
> those N places and re-indent there. Note that those N places can be
> different files.
>

Hmm.... why don't you simply try an indent-buffer, ...?

http://emacsblog.org/2007/01/17/indent-whole-buffer/

If I have a large amount of stuff I need to reindent I either select
and indent-region, or write some elisp to hack things up as needed..

kris

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Romain Bardou]

Le 21/09/2012 22:11, Kristopher Micinski a écrit :
> On Fri, Sep 21, 2012 at 9:42 AM, Romain Bardou<romain.bardou@inria.fr>  wrote:
>>> a configuration variable to change this behavior.  (BTW, do you know
>>> that M-q re-indent the current expression, so re-indenting after a name
>>> change is effortless.)
>>
>>
>> I know about M-q but I disagree that it is effortless: if you change a
>> variable name, and this name is used in N places, you have to go to all
>> those N places and re-indent there. Note that those N places can be
>> different files.
>>
>
> Hmm.... why don't you simply try an indent-buffer, ...?
>
> http://emacsblog.org/2007/01/17/indent-whole-buffer/
>
> If I have a large amount of stuff I need to reindent I either select
> and indent-region, or write some elisp to hack things up as needed..
>
> kris

Thank you for your input, Kristopher.

One big issue I have with indent-buffer (and also with M-q, actually) is 
that it will reindent stuff I have indented by hand to bypass some rules 
of Tuareg I don't like.

Even if I had an "indent correctly and instantly all files of my 
project" button, I still would not find the current indentation rule for 
application acceptable. I believe that good indentation rules should not 
rely on automation to actually be usable.

I'm not fluent at all with elisp and that's not something I am really 
interested in learning.

That's just personal taste really though.

Cheers,

-- 
Romain Bardou

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Dmitry Grebeniuk]

Hello.

> that it will reindent stuff I have indented by hand to bypass some rules of
> Tuareg I don't like.

> I believe that good indentation rules should not rely on
> automation to actually be usable.

  This is the common problem of all IDE's for OCaml I've tried,
strangely.
  For example, TypeRex, good IDE, lacks an option to disable
autoindentation, so I doubt whether I'll be fighting with
autoindentation or will stop to use TypeRex at all in some
near future.  (of course I filed a github issue, it was 2 months ago,
https://github.com/OCamlPro/typerex/issues/58
, still no reaction.)
  I don't want to learn elisp to fix this, too.

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Kristopher Micinski]

On Sat, Sep 22, 2012 at 9:22 AM, Romain Bardou <romain.bardou@inria.fr>
> Thank you for your input, Kristopher.
>
> One big issue I have with indent-buffer (and also with M-q, actually) is
> that it will reindent stuff I have indented by hand to bypass some rules of
> Tuareg I don't like.
>
> Even if I had an "indent correctly and instantly all files of my project"
> button, I still would not find the current indentation rule for application
> acceptable. I believe that good indentation rules should not rely on
> automation to actually be usable.
>
> I'm not fluent at all with elisp and that's not something I am really
> interested in learning.
>
> That's just personal taste really though.
>
> Cheers,

I completely understand that: indentation is something that requires
human intelligence (and good indentation is an art rather than
science).

For the same reasons I do not indent a whole buffer at once, usually..

FYI some open source projects have as part of their build process a
call to GNU indent, super annoying when indentation is not as you
describe..

That being said, if there is a part of tuareg mode that you need to
modify to accommodate your needs (the change of a variable name, as
you mention), hacking on the elisp might be preferable.  And yes, I
understand emacs lisp is this somewhat scary and involved language,
more for reasons related to emacs (as an API) rather than anything
having to do with lisp.. :-/

kris

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Edgar Friendly]

On 9/22/2012 2:13 PM, Kristopher Micinski wrote:
> ... indentation is something that requires
> human intelligence (and good indentation is an art rather than
> science).
I admit that under some circumstances, code can be made easier to read 
by giving it crafted indentation. But having a single indentation 
standard (and probably code layout) for all OCaml code would be better.  
In exchange for the occasional inability to indent code for extra 
readability (maybe comments would suffice instead), we would gain a 
community-wide improvement in our ability to read each others' code.

E.

Re: [Caml-list] Tuareg mode: Indentation of application arguments

[Kristopher Micinski]

On Sat, Sep 22, 2012 at 8:27 PM, Edgar Friendly <thelema314@gmail.com> wrote:
> On 9/22/2012 2:13 PM, Kristopher Micinski wrote:
>>
>> ... indentation is something that requires
>>
>> human intelligence (and good indentation is an art rather than
>> science).
>
> I admit that under some circumstances, code can be made easier to read by
> giving it crafted indentation. But having a single indentation standard (and
> probably code layout) for all OCaml code would be better.  In exchange for
> the occasional inability to indent code for extra readability (maybe
> comments would suffice instead), we would gain a community-wide improvement
> in our ability to read each others' code.
>
> E.
>

To be honest, most of the time it doesn't matter, because whoever you
work for will enforce a coding standard...

kris

[Caml-list] Re: Tuareg mode: Indentation of application arguments

[Stefan Monnier]

> One big issue I have with indent-buffer (and also with M-q, actually) is
> that it will reindent stuff I have indented by hand to bypass some rules of
> Tuareg I don't like.

If the places where you need to override Tuareg's indentation are very
few, and if you use Tuareg's SMIE indenter, then you might be able to
use the (* fixindent *) comment, which tells Tuareg to not mess with
this line's indentation.

Lots of "if"s, I know.  But for now, it's best I could come up with.

Of course, maybe the particular indentation you want to use could be
mechanized as well, so you might want to report the "poorly indented"
cases as bugs.


        Stefan