By the way, does Hat - the Haskell Tracer?

On Wed, Jul 15, 2009 at 11:13 PM, Don Stewart wrote:

fernanbolando:

...

Hi all

I recently used 2 hours of work looking for a bug that was causing

Program error: Prelude.!!: index too large

This is not very informative. It did not give me a hint which function was causing this. In C adding a few printf would have helped me, but in haskell I was not sure how to do that. Can anybody point me to some debuggin method everyone uses.

You could:

   * use Debug.Trace.trace  (equivalent of printf debugging)    * use asserts: the 'assert' function    * use the GHCi debugger to construct a stack trace    * use profiling to construct a stack trace.    * use the GHC head branch for first class stack traces, described          in, "Finding the needle: Stack Traces for GHC"                http://pubs.doc.ic.ac.uk/finding-the-needle/finding-the-needle.pdf

Another option would be to use Safe.atMay[1] instead of !!. [1]: http://hackage.haskell.org/packages/archive/safe/0.2/doc/html/Safe.html#v%3A... -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe

2009/7/16 Marc Weber

I recall there was another method. Yeah, I even found it (using ghci and set -fbreak-on-exception)

http://donsbot.wordpress.com/2007/11/14/no-more-exceptions-debugging-haskell...

Careful, better use -fbreak-on-error rather than -fbreak-on-exception. Some functions, e.g. 'doesFileExist', use exceptions internally and you might end up stopping at an exception that is benign. (Identifying the exception has become a little harder since we got extensible exceptions.) -- Push the envelope. Watch it bend.

Fernan Bolando writes:

On Thu, Jul 16, 2009 at 4:10 PM, Jon Fairbairn wrote:

...

I wonder if your code has to use !! at all? I took a look at a random module from the above link, and (without making much attempt at understanding it), I'd guess that using accumArray and friends would be more appropriate. Mostly you don't want to be doing indexing on lists.

Most of the !! are used to access a list of parameters similar to accessing structures in C.

Ah. I wondered why so many of them were x!!0. list!!constant is almost always something that could be done better.

I am not particularly sure how to apply accumArray for something like that?

You [probably] don't. My apologies for not looking more carefully (I really did just glance at it), but the remark about not using !! still stands. I've been trying to look at your source again, but the link doesn't respond to requests at the moment, so I can't be much help. Haskell has much richer mechanisms for doing structure-like things than C. Tuples and data types being the most obvious. -- Jón Fairbairn Jon.Fairbairn@cl.cam.ac.uk

Henning Thielemann writes:

On Thu, 16 Jul 2009, Fernan Bolando wrote:

...

Hi all

I recently used 2 hours of work looking for a bug that was causing

Program error: Prelude.!!: index too large

A good way to avoid such problems is to avoid partial functions at all. (!!) is also inefficient. Is it possible to define the function in terms of foldl?

I've looked at the code a bit more, and, with apologies to the original poster, it doesn't look much like Haskell. For example, in http://plan9.bell-labs.com/sources/contrib/fernan/escomma/Circuit.hs there's stuff beginning with tADM :: Int tADM = 1 tVSRC :: Int tVSRC = 2 tISRC :: Int tISRC = 3 ... that I think probably should be data Something = ADM | VSRC | ISRC ... deriving (Enum, ...) though when I get to "(((fst z0)!!pMSET)!!pMTYPE) == mOP" I'm at a loss to determine quite what the intention is. Maybe it should be an array indexed by enum types, maybe a function, maybe something that can be pattern matched on? I don't know. -- Jón Fairbairn Jon.Fairbairn@cl.cam.ac.uk

Fernan Bolando wrote:

The intention is z0 is a system parameter and database, it contains a set of info needed to define a particular simulation

it looks like ( [n,m...], [m,o,p])

n is is a list info settings for the circuit analysis m is a list of statistics for the circuits that is need in sub-sequent calculation m is a list of circuit settings like temperature etc. o is a list of matrix solution for non-linear newton raphson p is a list of matrix solution for time dependent calculations

This would be better as, data SystemParameterDatabase = SystemParameterDatabase { infoSettings :: InfoSettings , statistics :: Statistics , settings :: Settings , nlnr :: [Matrix Double] , timeDependent :: [Matrix Double] } data InfoSettings = InfoSettings { pMSET :: MSET , aSetting :: A , bSetting :: B ... } data Statistics = Statistics { aStatistic :: A , anotherStatistic :: A , bStatistic :: B ... } data Settings = Settings { temperature :: Kelvin , etc :: Etc } ... A single-constructor ADT, especially with the labeled-fields syntax, is pretty close to C structs; no need to reinvent them and give yourself headaches. Really, the only thing you should be using lists for is a variable-length sequence of elements drawn from the same type and distinguished only by their position in the sequence. Whenever the length is fixed or the (semantic) type of elements can be distinguished from one another (or altered by pushing/popping the list), then a list is not what you want to be using because it doesn't capture the intention of the type (and therefore allows unintelligible values of the type, and inappropriate transformations on the type). -- Live well, ~wren