June 2011 - web-devel

http-enumerator connecting via a http proxy
by Erik de Castro Lopo 25 Jun '11

25 Jun '11

Hi all, Has anyone managed to get http-enumerator connecting via a http proxy? I've tried constructing a custom 'Request m' object and submitting that the httpLbsRedirect but all I've managed to get is: connect: does not exist (Connection refused) Clues? Would this need a new http-enumerator function something like: httpProxyLbsRedirect :: (MonadIO m, Failure HttpException m) => String -> Int -> Request m -> Manager -> m Response where the first two parameters are the proxy hostname and port? Cheers, Erik -- ---------------------------------------------------------------------- Erik de Castro Lopo http://www.mega-nerd.com/

2 11

Some Fun: Panther AJAX Library
by Blake Rain 25 Jun '11

25 Jun '11

Dear All, Some history: About two or so years ago, before I'd heard of Yesod, we started a web development framework for Haskell which we called Panther. This framework was quite similar to ASP.Net or PHP: pages were source files and it had a pre-processor which generated Haskell code. To cut a long story short: I never really finished off Panther, as I found out about Michael's wonderful Yesod. I'm now resurrecting the project with a different goal: hosting multiple WAI applications over clusters. The fun part of the old Panther was that it used an evolutionary rewiring algorithm to distribute applications over clusters of computers over multiple sites. As you can imagine, that was a great joy to write :) When the newer Panther actually stabilizes (or even works) I'll explain various bits of it in more detail. One of the nice features that Panther had was the ability to specify client-side operations on the server. For example, given the HTML: > <div id="myDiv"> > <i>Some Text</i> > </div> > <input type="text" id="name" /> > <button type="button" panther:onclick="changeMyDiv"> > Change! > </button> We could replace the text in the <div> with "Other Text", change the style of the <div> and print the contents of the input element with the following Haskell code: > changeMyDiv :: (Monad m) => Element -> Ajax m () > changeMyDiv button = do > [div, input] <- getElementsById ["myDiv", "name"] > > setInnerHtml div "Other Text" > setStyle div ["font-weight" =: "bold", "color" =: "green"] > > v <- getValue input > liftIO $ print v (the code was actually a little more complex, as 'getElementsById' yielded a list of 'Maybe Element'. But you get the idea). In earlier days - before we adopted Yesod and rewrote everything :) - we found this functionality to be quite useful. We mostly used it to handle dynamic forms, progress reporting and so on. I'd completely forgotten about the panther-ajax package. The past few discussions on web-devel regarding dynamic forms made me think about how we have solved similar problems, and what we used to do in the past. Anyway, last week I decided to port the package over to Yesod and make some improvements to the API. I've put an experimental version on GitHub for those who are at all interested in this: https://github.com/MassiveTactical/panther-ajax There is a simple example in the repo. For those of you who don't want to build the example, but would still like see it working, the example is currently running at: http://meadowstalk.com:3141/ It should still be running... You can get an immediate idea for how it works by examining the network activity in Chrome (or whatever your browser may be). If any of you are interested - or are idly curious and have some spare time - I'd love to hear any feedback. The library still needs a lot of improvement. There is still quite a bit of code to port over, and the optimizing assembler needs some tender loving care before it can be used again. Kind Regards, Blake Rain Massive Tactical

2 1

Multi-selects for yesod-form in, let's keep rolling
by Michael Snoyman 24 Jun '11

24 Jun '11

Hi all, David McBride has implemented multi-selects in yesod-form (way to go!). This involves a breaking change to the API. So while we're already breaking things, I wanted to get some discuss dynamically-increasing forms in a more concrete way. I've been playing around with a possible API, and came up with something usable. You can see the result in Github: https://github.com/snoyberg/yesod-form/blob/master/Yesod/Form/MassInput.hs . You can also see a usage example in the getMassR handler in the hello-forms demo: https://github.com/snoyberg/yesod-form/blob/master/hello-forms.hs For now, the basic idea is that you'll provide inputList with a label for this set of fields, some way to convert the list of FieldViews generated by the individual values into a widget, and a widget for a single view. inputList will then create a single field in the larger form, the will contain within it multiple subforms. Each subform will have an option to delete it in addition to its normal contents. At the end of the list of subforms, there is a checkbox to add an extra row. My first use case for this is for creating pages in the Yesod wiki, and I *think* it will work well for that case (though I haven't tried it out yet to be sure). I wanted to get input from others about this, since this is a complicated feature that could go in many different directions. One thing to point out is that this API only addresses applicative forms, not monadic forms. Michael

1 0

Form Composition
by Michael Snoyman 24 Jun '11

24 Jun '11

Just passing this along: ---------- Forwarded message ---------- From: MightyByte <mightybyte(a)gmail.com> To: web-devel(a)haskell.org Date: Tue, 21 Jun 2011 12:47:07 -0400 Subject: Form Composition I wrote something like this for formlets[1] quite some time ago. (We called it massInput.) For all practical purposes, formlets has been superceded by digestive-functors [2]. I re-implemented it again for digest-functors (called inputList) with more features and better design[3]. It's a tricky problem, and I went down several dead-end paths before converging on the solution as it exists in digestive-functors now. If you're going to re-implement it with your form framework, I would highly recommend looking at our existing work first. [1] https://github.com/chriseidhof/formlets [2] https://github.com/jaspervdj/digestive-functors [3] http://hackage.haskell.org/packages/archive/digestive-functors/0.1.0.0/doc/… On Thur, June 16, 2011 at 6:43 PM, Michael Snoyman wrote: > This is definitely something missing from the yesod-form API as is. > I'm not quite sure how the API would look for this; has anyone given > this some thought? > > On Wed, Jun 15, 2011 at 7:06 AM, Tyson Whitehead <twhitehead@...> wrote: > > Hi All, > > > > I'm wanting to make a form for "[a]" by composing forms for "a" along with > > added a delete options and a final add another option. > > > > Is there an accepted way to do this sort of thing? It doesn't seem to me to > > be a good fit for the applicative framework as it is not a fixed structure. > > > > Cheers! -Tyson

4 3

Yesod form composition
by Tyson Whitehead 23 Jun '11

23 Jun '11

Hi All, I'm wanting to make a form for "[a]" by composing forms for "a" along with added a delete options and a final add another option. Is there an accepted way to do this sort of thing? It doesn't seem to me to be a good fit for the applicative framework as it is not a fixed structure. Cheers! -Tyson

3 4

Yesod Limitation in Route Methods?
by Eric Schug 23 Jun '11

23 Jun '11

I am trying to develop two similar route one that takes a constant string value as its last argument and one that takes a integer as its last argument /admin/#String/add ModelAddR GET /admin/#String/#Int ModelEditR GET I would expect that there should be no conflict and that the first should take precedence. However I get the following error. Exception when trying to run compile-time code: Overlapping routes: (Resource "ModelAddR" [StaticPiece "admin",SinglePiece "String",StaticPiece "add"] ["GET"],Resource "ModelEditR" [StaticPiece "admin",SinglePiece "String",SinglePiece "Int"] ["GET"]) Is this a limitation of Yesod or a bug? The following seems to work fine. /admin/#String/add ModelAddR GET /admin/#String/edit/#Int ModelEditR GET

1 0

confusion over yesod-forms-0.2.0
by Michael Litchard 22 Jun '11

22 Jun '11

I installed the new yesod-forms-0.2.0 I tried out the sample program hello-forms.hs. The new changes work just fine. No problems. I don't understand some of the notation, but that didn't stop me from using yesod-form prior to the changes. I copied the style of hello-form exactly, I think. But I got a compile error. I don't think the error indicates the actual problem, and I suspect I have left out something. Below is a link to hello-form, the code I am trying to compile and the error I get. I would appreciate any insight offered. hello-forms.hs http://hpaste.org/48177 Handler/SSLvpn.hs http://hpaste.org/48178 error Handler/SSLvpn.hs:43:14: No instance for (RenderMessage Aframe FormMessage) arising from a use of `selectField' Possible fix: add an instance declaration for (RenderMessage Aframe FormMessage) In the first argument of `areq', namely `(selectField tests)' In the second argument of `(<*>)', namely `areq (selectField tests) "Pick a test" Nothing' In the second argument of `($)', namely `pure (,) <*> areq (multiSelectField servers) "Pick some servers" Nothing <*> areq (selectField tests) "Pick a test" Nothing'

1 1

Re: [web-devel] yesod dependency hell
by Michael Litchard 22 Jun '11

22 Jun '11

My thinking is that this problem comes up anytime someone wants to play around with experimental code. As evidenced by the yesodwiki.cabal file, I think there's a more straightforward solution. I'm just not seeing it. On Wed, Jun 22, 2011 at 1:46 PM, Thomas Hartman <thomashartman1(a)googlemail.com> wrote: > one idea: move .ghc and .cabal to temp directory. > > start over with a clean slate. > > On Wed, Jun 22, 2011 at 1:30 PM, Michael Litchard <michael(a)schmong.org> wrote: >> Configuring Aframe-0.0.1... >> Warning: This package indirectly depends on multiple versions of the same >> package. This is highly likely to cause a compile failure. >> package yesod-auth-0.4.0.2 requires yesod-form-0.1.0.1 >> package yesod-0.8.2.1 requires yesod-form-0.1.0.1 >> package Aframe-0.0.1 requires yesod-form-0.2.0 >> >> I specified in Aframe.cabal yesod-form 0.2.0. The above warning is >> what I was looking to avoid. What is the solution to this problem? I >> could not find an obvious solution in yesodwiki.cabal, even though I >> see it's doing something similar to what I want. How did you deal with >> the above warning message? >> >> On Wed, Jun 22, 2011 at 12:32 PM, Michael Snoyman <michael(a)snoyman.com> wrote: >>> Have a look at the yesodwiki[1]; basically, you can't use the yesod >>> package itself, only the subpackages. This is by design so that a >>> single Yesod release will keep API stability, while users can still >>> experiment with newer versions of underlying packages. >>> >>> Michael >>> >>> [1] https://github.com/snoyberg/yesodwiki/blob/master/yesodwiki.cabal >>> >>> On Wed, Jun 22, 2011 at 10:29 PM, Michael Litchard <michael(a)schmong.org> wrote: >>>> I need to use yesod-form 0.2.0. The problem is that the yesod package, >>>> and the yesod-auth package expect yesod-form to be 0.1.0* >>>> I tried changing the .cabal files but that breaks in ways that makes >>>> me think the fix is not so simple. How do I go about being able to use >>>> yesod-form 0.2.0 >>>> >>>> _______________________________________________ >>>> web-devel mailing list >>>> web-devel(a)haskell.org >>>> http://www.haskell.org/mailman/listinfo/web-devel >>>> >>> >> >> _______________________________________________ >> web-devel mailing list >> web-devel(a)haskell.org >> http://www.haskell.org/mailman/listinfo/web-devel >> >

1 2

Re: [web-devel] [Haskell-cafe] [ANN] mysql-simple - your go-to package for talking to MySQL
by Bryan O'Sullivan 21 Jun '11

21 Jun '11

On Tue, Jun 21, 2011 at 7:47 AM, David Virebayre <dav.vire+haskell(a)gmail.com > wrote: > The problem isn't with the stored procedure, it works if I call it > from the mysql client. > Right - as I mentioned in my previous note, the problem is that stored procedures and multi-statement queries can both return multiple result sets. We can't easily use type inference to express the difference between "in this use of query, I want a single result" (the common case) and "in this other use of query, I expect three results, each with different shapes" (far less common), so we need something like a multiQuery function (and perhaps a MultiResult class) instead. Another unrelated thing : the documentation states that the Query type > is designed to make it difficult to create queries by concatenating > strings. > You can do it, but you have to use the Monoid class's functions, e.g.: "select " `mappend` "2 + 2" For cases like your "show columns from" example, though, I prefer Chris's suggestion of creating a custom newtype with its own special Param instance.

1 0

Re: [web-devel] [Haskell-cafe] [ANN] mysql-simple - your go-to package for talking to MySQL
by Bryan O'Sullivan 21 Jun '11

21 Jun '11

On Tue, Jun 21, 2011 at 4:45 AM, David Virebayre <dav.vire+haskell(a)gmail.com > wrote: > > I had trouble accessing the documentation : the last versions on > hackage have a build failure, so the doc isn't available. I don't understand why that build failure occurs. You can always build documentation for a package locally using "cabal haddock". > The very first example didn't work for me : Oops, thanks for mentioning that. I've fixed the documentation and made the information about type inference a little clearer. Indeed, the documentation shows that convertError takes 3 parameters, and I gave, as per the example, only 2. > But I'm not sure what to write for the 3rd parameter, the > documentation doesn't help me here. > Thanks for spotting the omission, I've clarified that. The Int parameter indicates the number of columns expected for conversion. > I'm not sure if that means Database.MySQL supports calling stored > procedures that return a result set or not. I suspect not. The mysql-simple package currently doesn't support stored procedures or multi-statement queries. Each of these can return multiple result sets. It would be necessary to add new functions to the API to support them, as of course each result set could have a different shape.

1 0