Re: [Haskell-cafe] Haskell and the Software design process

newer
ANN: test-framework 0.3.0

older
Haskell interface to Frama-C

Henning Thielemann

3 May 2010 3 May '10

11:21 a.m.

On Mon, 3 May 2010, Gregory Collins wrote:

...

Don Stewart writes:

...
Some key points:

* Avoid partial functions

As an important corollary to this one I would add: "never throw exceptions from pure code". They often leak out from "catch" blocks and ruin your day.

It's not possible to throw exceptions from pure code. You can only call 'error' and that's another name for 'undefined', i.e. you have a partial (non-total ?) function.

Show replies by date

Gregory Collins

3 May 3 May

11:27 a.m.

New subject: Haskell and the Software design process

Henning Thielemann writes:

...

It's not possible to throw exceptions from pure code. You can only call 'error' and that's another name for 'undefined', i.e. you have a partial (non-total ?) function.

http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E... ? G. -- Gregory Collins

Henning Thielemann

11:34 a.m.

New subject: Haskell and the Software design process

Gregory Collins schrieb:

...

Henning Thielemann writes:

...
It's not possible to throw exceptions from pure code. You can only call 'error' and that's another name for 'undefined', i.e. you have a partial (non-total ?) function.

http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E...

I see. This should be forbidden, at all! :-)

Ketil Malde

12:04 p.m.

New subject: Haskell and the Software design process

Henning Thielemann writes:

...

...
http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E...

...

I see. This should be forbidden, at all! :-)

Why is this worse than or different from 'error'? To me it looks like 'error', only with a non-string parameter. -k -- If I haven't seen further, it is by standing in the footprints of giants

Henning Thielemann

12:14 p.m.

New subject: Haskell and the Software design process

Ketil Malde schrieb:

...

Henning Thielemann writes:

...
...
http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E...

...
I see. This should be forbidden, at all! :-)

Why is this worse than or different from 'error'? To me it looks like 'error', only with a non-string parameter.

Because it encourages the use in a non-error way, that is, catching such an exception is regarded as good use.

Brandon S. Allbery KF8NH

7 May 7 May

9:03 p.m.

New subject: Haskell and the Software design process

On May 3, 2010, at 12:14 , Henning Thielemann wrote:

...

Ketil Malde schrieb:

...
Henning Thielemann writes:

...
...
http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E...

...
I see. This should be forbidden, at all! :-)

Why is this worse than or different from 'error'? To me it looks like 'error', only with a non-string parameter.

Because it encourages the use in a non-error way, that is, catching such an exception is regarded as good use.

One could argue that it's a more comprehensible to mortals version of Control.Monad.Cont. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH

Gregory Collins

3 May 3 May

1:25 p.m.

New subject: Haskell and the Software design process

Ketil Malde writes:

...

Henning Thielemann writes:

...
...
http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E...

...
I see. This should be forbidden, at all! :-)

Why is this worse than or different from 'error'? To me it looks like 'error', only with a non-string parameter.

Calling 'error' is just as bad IMO! G -- Gregory Collins

Ketil Malde

4:55 p.m.

New subject: Haskell and the Software design process

Gregory Collins writes:

...

...
Henning Thielemann writes:

...

...
...
...
http://www.haskell.org/ghc/docs/6.12.2/html/libraries/base-4.2.0.1/Control-E...

...

...
...
I see. This should be forbidden, at all! :-)

...

...
Why is this worse than or different from 'error'? To me it looks like 'error', only with a non-string parameter.

...

Calling 'error' is just as bad IMO!

Which is the same as 'as good'? I think error is legitimate in many cases. For instance: * when you know a branch will never be taken, you can add an 'error "this will never happen"' which both informs other readers of the code of the intended usage, and provides a greppable message if you were wrong. * for stub functions during development ('undefined' doesn't tell you which undefined value you tried to evaluate) * Functions like division, which is still useful, even if it is partial. -k -- If I haven't seen further, it is by standing in the footprints of giants

5493

Age (days ago)

5498

Last active (days ago)

List overview

Download

7 comments

5 participants

participants (5)

Brandon S. Allbery KF8NH
Gregory Collins
Henning Thielemann
Henning Thielemann
Ketil Malde

Re: [Haskell-cafe] Haskell and the Software design process

tags

participants (5)