Comments on: Indentation-sensitive Erlang

By: Ulf Wiger

Ulf Wiger — Mon, 16 Feb 2009 08:36:54 +0000

@Orielt – Could you expand on what it is that doesn’t work? I just checked it with Safari/Windows, and it wasn’t immediately obvious to me what would be wrong (posting this comment from Safari, btw).

By: OrieltOutweig

OrieltOutweig — Sun, 15 Feb 2009 21:26:27 +0000

Your site does not correctly work in safari browser

By: Bill McKeeman

Bill McKeeman — Tue, 30 Sep 2008 13:33:49 +0000

Python required indentation has some value for the final code, and does avoid the need for end-brackets. I have found it both irritating and dangerous in refactoring. It is all too easy to move some code and have it accidentally syntactically correct when it should be indented or outdented. Now the code is pretty and wrong. I prefer pretty printers.

By: Arthur Lee

Arthur Lee — Tue, 29 Apr 2008 02:23:31 +0000

@thomas lackner:
Using tab char should definitely be avoided. Just use normal spaces for indentations.

By: Ulf Wiger

Ulf Wiger — Thu, 20 Mar 2008 09:44:05 +0000

Oh, yes, here was another problem:
Alternative function clauses are separated by ‘;’, which with indentation could be replaced by an ‘END’. But if ‘END’ terminates the last function clause, we get a syntax error, unless we change the grammar in a way similar as we would if we wanted every function clause to end with a ‘dot’.

So what’s the solution? Perhaps we should have a token denoting an empty line? Then, successive function clauses could be separated by newlines (without ‘;’), and the end of the function is marked by an empty line?

By: Ulf Wiger

Ulf Wiger — Thu, 20 Mar 2008 08:01:40 +0000

I did try to omit the ‘end’, but were unsuccessful in my first attempt. Perhaps there is a clever way to formulate the grammar so that it works, but I have yet to spot it. I welcome any help in the matter.

The problem with getting rid of the ‘dot’ was that the erlang compiler scans one form at a time, and the form ends with a dot. I guess the scanner could signal end-of-form at either a ‘dot’ or an ‘END’, whichever comes first…

By: carl henrik

carl henrik — Wed, 19 Mar 2008 20:54:36 +0000

“The ending dots don’t have to be on their own line. Getting rid of them was too hard for me, since ‘dot’ is the end token for the Erlang grammar.”

but it seems so easy; you’d need to generate an END token when indentation goes to zero, or @ eof?
what am i missing?

By: Chris Okasaki

Chris Okasaki — Wed, 19 Mar 2008 19:19:17 +0000

Cool experiment!

“(Note especially that the final ‘end’ must be aligned with the Y, rather than the ‘case’. Perhaps this could be avoided…?)”

A fairly typical approach would be to use the indentation to allow you to omit the ‘end’ altogether.

By: Paul Brown

Paul Brown — Wed, 19 Mar 2008 19:06:28 +0000

Reads a lot like Haskell now… At least with Emacs as a crutch, the Haskell’s indentation rules do help with the visual detection of errors.

By: thomas lackner

thomas lackner — Wed, 19 Mar 2008 18:34:58 +0000

Interesting hack! I loathe indentation-based syntax (after all, the tab is an invisible character.. not a wise denotation for important structure), but it’s really cool that you were able to do this without a complete rewrite of the parser. Is it possible that we could use a similar trick to add other useful tricks to erlang, like a syntax-level records fix, or literal syntax like { name: “Tom”, age: 26 } for proplists?