do-and-if-then-else modification - Haskell-prime - Haskell.org

newer
Polymorphic components, so far

do-and-if-then-else modification

older
Fixity resolution, possible...

isaac jones

17 Feb 2007 17 Feb '07

12:44 a.m.

Iavor and I just made the trivial modification for DoAndIfThenElse syntax as described here: http://hackage.haskell.org/trac/haskell-prime/wiki/DoAndIfThenElse You can see the change here: http://darcs.haskell.org/haskell-prime-report/report/haskell-report-html/exp... (Is there anything else that needs to change?) As always, we provide a quick view of the status of Haskell' here: http://hackage.haskell.org/trac/haskell-prime/wiki/Status%27 Any comments on this modification? How do people feel about the suggestion that we do it for case statements as well? peace, isaac

Reply

Sign in to reply online Use email software

Show replies by date

Benjamin Franksen

18 Feb 18 Feb

7:19 p.m.

isaac jones wrote:

Iavor and I just made the trivial modification for DoAndIfThenElse syntax as described here: http://hackage.haskell.org/trac/haskell-prime/wiki/DoAndIfThenElse

You can see the change here:

http://darcs.haskell.org/haskell-prime-report/report/haskell-report-html/exp...

(Is there anything else that needs to change?)

"Section 3.6 Conditionals" would have to be changed accordingly. It still says "exp -> if exp1 then exp2 else exp3". Ben

Reply

Sign in to reply online Use email software

Iavor Diatchki

19 Feb 19 Feb

5:51 p.m.

Hello, On 2/18/07, Benjamin Franksen wrote:

"Section 3.6 Conditionals" would have to be changed accordingly. It still says "exp -> if exp1 then exp2 else exp3". Thanks! I fixed this too.

Just for the record, I don't think that this change is particularly useful at all. We made the change because it is on the "definitely in" list accepted by the Haskell' committee and it seemed quite easy to do. -Iavor

Reply

Sign in to reply online Use email software

Malcolm Wallace

11:16 a.m.

isaac jones wrote:

Iavor and I just made the trivial modification for DoAndIfThenElse

Any comments on this modification? How do people feel about the suggestion that we do it for case statements as well?

I'm happy to allow the change for if-then-else. The need for extra indentation there has annoyed me a few times. However, I have never come across the case-of expression variation in practice. Indeed, the wiki page is not very clear on what exactly the problem is. For instance, until a few minutes ago, the given example was merely: case whatever of blah -> bluh blah' -> bluh' However, that is already perfectly legal Haskell'98. I just changed the example to read do case whatever of blah -> bluh blah' -> bluh' since that is what I thought was perhaps intended. And indeed that is rejected by a H'98 parser. But it seems a bit of an unlikely coding style. And it now occurs to me that perhaps something different was intended. The following is legal H'98: case whatever of Cons -> bluh Nil -> foo But stick it in a do block and it ceases to be accepted: do case whatever of Cons -> bluh Nil -> foo Now for myself, I tend to strictly indent the alternatives in a case expression, but I can see that if you prefer non-indented alternatives, then the latter would indeed be annoying. Regards, Malcolm

Reply

Sign in to reply online Use email software

Ross Paterson

11:57 a.m.

On Mon, Feb 19, 2007 at 11:16:03AM +0000, Malcolm Wallace wrote:

isaac jones wrote:

...
Iavor and I just made the trivial modification for DoAndIfThenElse

Any comments on this modification? How do people feel about the suggestion that we do it for case statements as well?

I'm happy to allow the change for if-then-else. The need for extra indentation there has annoyed me a few times.

The semicolon before the "else" is the most useful. Adding another before the "then" has less motivation.

However, I have never come across the case-of expression variation in practice. Indeed, the wiki page is not very clear on what exactly the problem is.

The argument regarding "case" is that if you add a semicolon before "then", it would be consistent to also add one before "of".

Reply

Sign in to reply online Use email software

Malcolm Wallace

12:10 p.m.

Ross Paterson wrote:

The semicolon before the "else" is the most useful. Adding another before the "then" has less motivation.

do if foo then bar else baz seems just as likely to occur as do if foo then bar else baz particularly if the condition "foo" is large, and/or indentation is already running a long distance to the right.

...
However, I have never come across the case-of expression variation in practice. Indeed, the wiki page is not very clear on what exactly the problem is.

The argument regarding "case" is that if you add a semicolon before "then", it would be consistent to also add one before "of".

If the only argument is consistency, then surely the same applies to the "in" in a let expression. But of course this would correct a problem that never occurs in practice, because in a do-block one uses plain let, not let-in. So what evidence is there that case-of is a problem in practice? Regards, Malcolm

Reply

Sign in to reply online Use email software

Brian Smith

12:11 p.m.

On 2/19/07, Ross Paterson wrote:

On Mon, Feb 19, 2007 at 11:16:03AM +0000, Malcolm Wallace wrote:

...
isaac jones wrote:

...
Iavor and I just made the trivial modification for DoAndIfThenElse

Any comments on this modification? How do people feel about the suggestion that we do it for case statements as well?

I'm happy to allow the change for if-then-else. The need for extra indentation there has annoyed me a few times.

The semicolon before the "else" is the most useful. Adding another before the "then" has less motivation.

I think it is a bad idea to make the non-layout syntax more complicated and less readable, just so that, indirectly, the layout syntax can be improved. Improvements to the layout mechanism should be localized to the section of the report describing layout, and such improvements shouldn't do anything to harm the non-layout syntax. In particular, why you don't just say that "else" and "of" act like ";" during layout? do { q <- if x ; then y ; else z ; return q } - Brian

Reply

Sign in to reply online Use email software

Ross Paterson

12:32 p.m.

On Mon, Feb 19, 2007 at 07:11:24PM +0700, Brian Smith wrote:

I think it is a bad idea to make the non-layout syntax more complicated and less readable, just so that, indirectly, the layout syntax can be improved.

It doesn't: these semicolons are optional.

Reply

Sign in to reply online Use email software

Ross Paterson

12:53 p.m.

On Mon, Feb 19, 2007 at 11:16:03AM +0000, Malcolm Wallace wrote:

And it now occurs to me that perhaps something different was intended. The following is legal H'98:

case whatever of Cons -> bluh Nil -> foo

But stick it in a do block and it ceases to be accepted:

do case whatever of Cons -> bluh Nil -> foo

No, that's a separate issue: NondecreasingIndentation.

Reply

Sign in to reply online Use email software

7032

Age (days ago)

7034

Last active (days ago)

Download

8 comments

6 participants

tags

participants (6)

Benjamin Franksen
Brian Smith
Iavor Diatchki
isaac jones
Malcolm Wallace
Ross Paterson