On 12 Mar 2009, at 15:33, Colin Paul Adams wrote:

...

...

...

...

...

"Thomas" == Thomas Davie writes:

Thomas> On 12 Mar 2009, at 15:16, Andrew Wagner wrote:

...

Can you expand on this a bit? I'm curious why you think this.

Thomas> For two reasons:

Thomas> Firstly, I often find that people use the Monadic Thomas> interface when one of the less powerful ones is both Thomas> powerful enough and more convenient, parsec is a wonderful Thomas> example of this. When the applicative instance is used Thomas> instead of the monadic one, programs rapidly become more Thomas> readable, because they stop describing the order in which Thomas> things should be parsed, and start describing the grammar Thomas> of the language being parsed instead.

That's interesting.

I recently used parsec 3. I wrote using the monadic interface because I could understand (just about) how to do so. I was looking at the examples in RWH, and I could follow the explanation of the monadic interface much easier. Perhaps this was because RWH shows how to write using the monadic interface, and then shows how to convert this to the applicative interface. It's hard to follow a tutorial that shows you how to convert from something you aren't starting with.

I suspect that this is an interesting corner case of both my two reasons – firstly, Monads are too powerful here, and secondly, you perhaps found it easier to think about the operational aspects of the parser than to think about the denotation of what the parser should parse. Is that somewhere accurate? Bob

Hmm, perhaps what we need is another monad tutorial that.....err...never mind. Actually, I'd like to see more written about Applicatives. I think they just finally clicked with me when reading the Typeclassopedia, and seeing the intended way to use them. Before it was always like "ok...so, if I've got a function already in my functor, I could use this, but...why would I have that?" On Thu, Mar 12, 2009 at 10:28 AM, Thomas Davie wrote:

On 12 Mar 2009, at 15:16, Andrew Wagner wrote:

Can you expand on this a bit? I'm curious why you think this.

...

For two reasons:

Firstly, I often find that people use the Monadic interface when one of the less powerful ones is both powerful enough and more convenient, parsec is a wonderful example of this. When the applicative instance is used instead of the monadic one, programs rapidly become more readable, because they stop describing the order in which things should be parsed, and start describing the grammar of the language being parsed instead.

Secondly, It seems relatively common now for beginners to be told about the IO monad, and start writing imperative code in it, and thinking that this is what Haskell programming is. I have no problem with people writing imperative code in Haskell, it's an excellent imperative language. However, beginners seeing this, and picking it up is usually counter productive – they never learn how to write things in a functional way, and miss out on most of the benefits of doing so.

Hope that clarifies what I meant :)

Bob

Conal,Do you think imperative Haskell can be a sort of "gateway drug" to "real" haskell? On Thu, Mar 12, 2009 at 11:25 AM, Conal Elliott wrote:

Thank you Bob! I'll throw in another 2 cents:

Yes, *one* aspect of Haskell is that it's a power tool for imperative programming -- a clever way to keep plugging away at the old sequential von Neumann paradigm. C++++.

I'd rather we strongly encourage Haskell-newbies toward shifting out of the imperative paradigm to thinking and programming *functionally*. It's a big shift, to make, and imperative-Haskell is a relatively easy substitute.

- Conal

On Thu, Mar 12, 2009 at 7:28 AM, Thomas Davie wrote:

...

On 12 Mar 2009, at 15:16, Andrew Wagner wrote:

Can you expand on this a bit? I'm curious why you think this.

...

For two reasons:

Firstly, I often find that people use the Monadic interface when one of the less powerful ones is both powerful enough and more convenient, parsec is a wonderful example of this. When the applicative instance is used instead of the monadic one, programs rapidly become more readable, because they stop describing the order in which things should be parsed, and start describing the grammar of the language being parsed instead.

Secondly, It seems relatively common now for beginners to be told about the IO monad, and start writing imperative code in it, and thinking that this is what Haskell programming is. I have no problem with people writing imperative code in Haskell, it's an excellent imperative language. However, beginners seeing this, and picking it up is usually counter productive – they never learn how to write things in a functional way, and miss out on most of the benefits of doing so.

Hope that clarifies what I meant :)

Bob_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Jeff Heard wrote:

Come to think of it, I've never seen an applicative tutorial of Parsec, only a monadic one. Does such a beast exist

Here's something that comes close: http://www.cs.uu.nl/research/techreps/repo/CS-2008/2008-044.pdf It's not specific to Parsec; instead, a parser type is defined on the fly, based on Philip Wadler's list of successes. The applicative and monadic approaches are clearly separated. Martijn.

I was thinking the same thing. If I remember correctly, RWH does a parser in an applicative style, but I'll have to look when I get home to be sure. If so, then maybe we could try doing the same thing in a monadic style, for comparison and contrast purposes. On Thu, Mar 12, 2009 at 11:29 AM, Jeff Heard wrote:

Come to think of it, I've never seen an applicative tutorial of Parsec, only a monadic one. Does such a beast exist, and if so, maybe we could merge the two together, work the same example in both, and thus help the programmer make the shift from monadic to applicative, from order of parsing to describing the grammar.

-- Jeff

...

Thank you Bob! I'll throw in another 2 cents:

Yes, *one* aspect of Haskell is that it's a power tool for imperative programming -- a clever way to keep plugging away at the old sequential von Neumann paradigm. C++++.

I'd rather we strongly encourage Haskell-newbies toward shifting out of

...

imperative paradigm to thinking and programming *functionally*. It's a big shift, to make, and imperative-Haskell is a relatively easy substitute.

- Conal

On Thu, Mar 12, 2009 at 7:28 AM, Thomas Davie wrote:

...

On 12 Mar 2009, at 15:16, Andrew Wagner wrote:

...

Can you expand on this a bit? I'm curious why you think this.

For two reasons:

Firstly, I often find that people use the Monadic interface when one of the less powerful ones is both powerful enough and more convenient,

2009/3/12 Conal Elliott : the parsec

...

...

is a wonderful example of this. When the applicative instance is used instead of the monadic one, programs rapidly become more readable, because they stop describing the order in which things should be parsed, and start describing the grammar of the language being parsed instead.

Secondly, It seems relatively common now for beginners to be told about the IO monad, and start writing imperative code in it, and thinking that this is what Haskell programming is. I have no problem with people writing imperative code in Haskell, it's an excellent imperative language. However, beginners seeing this, and picking it up is usually counter productive – they never learn how to write things in a functional way, and miss out on most of the benefits of doing so.

Hope that clarifies what I meant :)

Bob_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe