Haskell Platform Proposal: add the 'aeson' library - Libraries - Haskell.org

newer
Using lenses

Haskell Platform Proposal: add the 'aeson' library

older
OT: Symbolic type variables (was...

Bas van Dijk

17 Sep 2013 17 Sep '13

11:48 a.m.

Hello, I would like to propose adding the 'aeson' library to the next Haskell Platform which is scheduled for November. The detailed proposal is kept at: http://trac.haskell.org/haskell-platform/wiki/Proposals/aeson Discussion deadline: October 7th, 2013 Regards, Bas

Reply

Sign in to reply online Use email software

Show replies by date

Edward Kmett

17 Sep 17 Sep

2:53 p.m.

+1 from me. It is pretty much the defacto standard way to work with json these days. -Edward On Tue, Sep 17, 2013 at 11:48 AM, Bas van Dijk wrote:

Hello,

I would like to propose adding the 'aeson' library to the next Haskell Platform which is scheduled for November.

The detailed proposal is kept at:

http://trac.haskell.org/haskell-platform/wiki/Proposals/aeson

Discussion deadline: October 7th, 2013

Regards,

Bas _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Dan Burton

3:02 p.m.

+1 I agree with Edward. Everybody already uses, trusts, and loves aeson. -- Dan Burton On Tue, Sep 17, 2013 at 11:53 AM, Edward Kmett wrote:

+1 from me.

It is pretty much the defacto standard way to work with json these days.

-Edward

On Tue, Sep 17, 2013 at 11:48 AM, Bas van Dijk wrote:

...
Hello,

I would like to propose adding the 'aeson' library to the next Haskell Platform which is scheduled for November.

The detailed proposal is kept at:

http://trac.haskell.org/haskell-platform/wiki/Proposals/aeson

Discussion deadline: October 7th, 2013

Regards,

Bas _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Christopher Done

3:24 p.m.

+1 here. Standard JSON library. On 17 September 2013 21:02, Dan Burton wrote:

+1

I agree with Edward. Everybody already uses, trusts, and loves aeson.

-- Dan Burton

On Tue, Sep 17, 2013 at 11:53 AM, Edward Kmett wrote:

...
+1 from me.

It is pretty much the defacto standard way to work with json these days.

-Edward

On Tue, Sep 17, 2013 at 11:48 AM, Bas van Dijk wrote:

...
Hello,

I would like to propose adding the 'aeson' library to the next Haskell Platform which is scheduled for November.

The detailed proposal is kept at:

http://trac.haskell.org/haskell-platform/wiki/Proposals/aeson

Discussion deadline: October 7th, 2013

Regards,

Bas _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

David Luposchainsky

3:38 p.m.

+1

Reply

Sign in to reply online Use email software

Johan Tibell

3:59 p.m.

+1 On Tue, Sep 17, 2013 at 12:38 PM, David Luposchainsky wrote:

+1 _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

amindfv＠gmail.com

5:38 p.m.

+1, same reasons Tom El Sep 17, 2013, a las 15:59, Johan Tibell escribió:

+1

On Tue, Sep 17, 2013 at 12:38 PM, David Luposchainsky wrote:

...
+1 _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Alejandro Serrano Mena

5:45 p.m.

+1 2013/9/17

+1, same reasons

Tom

El Sep 17, 2013, a las 15:59, Johan Tibell escribió:

...
+1

On Tue, Sep 17, 2013 at 12:38 PM, David Luposchainsky wrote:

...
+1 _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Niklas Hambüchen

18 Sep 18 Sep

12:30 a.m.

On 18/09/13 04:02, Dan Burton wrote:

Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust: "Partial instances" that will call error in pure code and crash my program. ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it. https://github.com/bos/aeson/issues/126

Reply

Sign in to reply online Use email software

Bas van Dijk

3:02 a.m.

On 18 September 2013 06:30, Niklas Hambüchen wrote:

On 18/09/13 04:02, Dan Burton wrote:

...
Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust:

"Partial instances" that will call error in pure code and crash my program.

ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it.

https://github.com/bos/aeson/issues/126

What about settling that issue by adding documentation to the ToJSON ByteString instance and the 'encode' function warning users about this? Bas

Reply

Sign in to reply online Use email software

Simon Hengel

3:09 a.m.

On Wed, Sep 18, 2013 at 09:02:16AM +0200, Bas van Dijk wrote:

On 18 September 2013 06:30, Niklas Hambüchen wrote:

...
On 18/09/13 04:02, Dan Burton wrote:

...
Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust:

"Partial instances" that will call error in pure code and crash my program.

ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it.

https://github.com/bos/aeson/issues/126

What about settling that issue by adding documentation to the ToJSON ByteString instance and the 'encode' function warning users about this?

That looks scary. I'd either make it total or remove the instance all together. I think by now almost everybody in the Haskell community agrees that we do not want pure partial functions. Cheers, Simon

Reply

Sign in to reply online Use email software

Michael Snoyman

3:22 a.m.

On Wed, Sep 18, 2013 at 10:09 AM, Simon Hengel wrote:

...
On 18 September 2013 06:30, Niklas Hambüchen wrote:

...
On 18/09/13 04:02, Dan Burton wrote:

...
Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust:

"Partial instances" that will call error in pure code and crash my

On Wed, Sep 18, 2013 at 09:02:16AM +0200, Bas van Dijk wrote: program.

...
...
ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it.

https://github.com/bos/aeson/issues/126

What about settling that issue by adding documentation to the ToJSON ByteString instance and the 'encode' function warning users about this?

That looks scary. I'd either make it total or remove the instance all together.

I think by now almost everybody in the Haskell community agrees that we do not want pure partial functions.

I don't want this issue to block inclusion of aeson. And IMO, the HP review process shouldn't be a time to try and twist a maintainer's arm into making a change he/she isn't interested in. That happened quite a bit on the text package, and I don't want to see it happen on aeson as well (especially given that Bryan would be the victim in both cases). That said, I personally would prefer removing the ToJSON/FromJSON instances for ByteString. We've been bitten by this in our codebases, and have resorted to using a ByteString64 newtype wrapper, which uses base64 encoding to ensure safe roundtripping. So: +1 from me on including aeson as-is, and if Bryan's interested in making this change to aeson, I'd consider it a perk. Michael

Reply

Sign in to reply online Use email software

Herbert Valerio Riedel

5:24 a.m.

On 2013-09-18 at 09:22:07 +0200, Michael Snoyman wrote: [...]

So: +1 from me on including aeson as-is, and if Bryan's interested in making this change to aeson, I'd consider it a perk.

+1 like Micheal (i.e. I support `aeson` in HP unconditionally; but I'd be delighted if the `ByteString` instances not pure issue would be fixed somehow)

Reply

Sign in to reply online Use email software

Thomas Schilling

7:26 a.m.

I'm also +1, though I'd strongly prefer that the instance for ByteString gets dropped before inclusion. The "don't do that" approach is not very Haskelly. A Char8 ByteString is the same type as a Word8 ByteString, so while you could argue that Char8 is implicitly ASCII, the instance couldn't distinguish it from Word8 ByteStrings where this doesn't make sense. Since JSON doesn't have a standard way of encoding binary data (utf8/base64/base64url/...), external JSON API often may use their own convention, so it's best to require users to be explicit about semantics. On 18 September 2013 11:24, Herbert Valerio Riedel wrote:

On 2013-09-18 at 09:22:07 +0200, Michael Snoyman wrote:

[...]

...
So: +1 from me on including aeson as-is, and if Bryan's interested in making this change to aeson, I'd consider it a perk.

+1 like Micheal

(i.e. I support `aeson` in HP unconditionally; but I'd be delighted if the `ByteString` instances not pure issue would be fixed somehow) _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Bardur Arantsson

7:39 a.m.

On 2013-09-18 13:26, Thomas Schilling wrote:

I'm also +1, though I'd strongly prefer that the instance for ByteString gets dropped before inclusion.

The "don't do that" approach is not very Haskelly. A Char8 ByteString is the same type as a Word8 ByteString, so while you could argue that Char8 is implicitly ASCII, the instance couldn't distinguish it from Word8 ByteStrings where this doesn't make sense. Since JSON doesn't have a standard way of encoding binary data (utf8/base64/base64url/...), external JSON API often may use their own convention, so it's best to require users to be explicit about semantics.

+1 with removal of BS instance, -0 otherwise. (For all the reasons stated in this sub-thread.)

Reply

Sign in to reply online Use email software

Bardur Arantsson

7:41 a.m.

On 2013-09-18 13:39, Bardur Arantsson wrote:

On 2013-09-18 13:26, Thomas Schilling wrote:

...
I'm also +1, though I'd strongly prefer that the instance for ByteString gets dropped before inclusion.

...
The "don't do that" approach is not very Haskelly. A Char8 ByteString is the same type as a Word8 ByteString, so while you could argue that Char8 is implicitly ASCII, the instance couldn't distinguish it from Word8 ByteStrings where this doesn't make sense. Since JSON doesn't have a standard way of encoding binary data (utf8/base64/base64url/...), external JSON API often may use their own convention, so it's best to require users to be explicit about semantics.

+1 with removal of BS instance, -0 otherwise.

Sigh... pressed "Send" too fast. Should have said: +1 with removal of BS instance or explicit opt-in with appropriate documentation warnings (etc.). -0 otherwise

Reply

Sign in to reply online Use email software

John Lato

10:59 a.m.

I agree with Michael and Thomas, +1 to including aeson as-is; if the ByteString instance is removed so much the better. On Wed, Sep 18, 2013 at 6:26 AM, Thomas Schilling wrote:

I'm also +1, though I'd strongly prefer that the instance for ByteString gets dropped before inclusion.

The "don't do that" approach is not very Haskelly. A Char8 ByteString is the same type as a Word8 ByteString, so while you could argue that Char8 is implicitly ASCII, the instance couldn't distinguish it from Word8 ByteStrings where this doesn't make sense. Since JSON doesn't have a standard way of encoding binary data (utf8/base64/base64url/...), external JSON API often may use their own convention, so it's best to require users to be explicit about semantics.

On 18 September 2013 11:24, Herbert Valerio Riedel wrote:

...
On 2013-09-18 at 09:22:07 +0200, Michael Snoyman wrote:

[...]

...
So: +1 from me on including aeson as-is, and if Bryan's interested in making this change to aeson, I'd consider it a perk.

+1 like Micheal

(i.e. I support `aeson` in HP unconditionally; but I'd be delighted if the `ByteString` instances not pure issue would be fixed somehow) _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Edward Kmett

11:17 a.m.

I'm with Michael on +1 as-is. I'm rather neutral on removing the instance as it is a matter of taste and I tend to prefer these sorts of things to be simple up-down votes of confidence rather than have them devolve into long unproductive bikeshedding discussions. -Edward On Wed, Sep 18, 2013 at 3:22 AM, Michael Snoyman wrote:

On Wed, Sep 18, 2013 at 10:09 AM, Simon Hengel wrote:

...
...
On 18 September 2013 06:30, Niklas Hambüchen wrote:

...
On 18/09/13 04:02, Dan Burton wrote:

...
Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust:

"Partial instances" that will call error in pure code and crash my

On Wed, Sep 18, 2013 at 09:02:16AM +0200, Bas van Dijk wrote: program.

...
...
ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it.

https://github.com/bos/aeson/issues/126

What about settling that issue by adding documentation to the ToJSON ByteString instance and the 'encode' function warning users about this?

That looks scary. I'd either make it total or remove the instance all together.

I think by now almost everybody in the Haskell community agrees that we do not want pure partial functions.

I don't want this issue to block inclusion of aeson. And IMO, the HP review process shouldn't be a time to try and twist a maintainer's arm into making a change he/she isn't interested in. That happened quite a bit on the text package, and I don't want to see it happen on aeson as well (especially given that Bryan would be the victim in both cases).

That said, I personally would prefer removing the ToJSON/FromJSON instances for ByteString. We've been bitten by this in our codebases, and have resorted to using a ByteString64 newtype wrapper, which uses base64 encoding to ensure safe roundtripping.

So: +1 from me on including aeson as-is, and if Bryan's interested in making this change to aeson, I'd consider it a perk.

Michael

_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Reply

Sign in to reply online Use email software

Bas van Dijk

23 Sep 23 Sep

8:56 a.m.

On 18 September 2013 06:30, Niklas Hambüchen wrote:

On 18/09/13 04:02, Dan Burton wrote:

...
Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust:

"Partial instances" that will call error in pure code and crash my program.

ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it.

https://github.com/bos/aeson/issues/126 _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Note that Bryan pushed a patch which removes the ByteString instances: https://github.com/bos/aeson/commit/9ee73f303cacb8ae73b2325f2f47288522607420 I think this is a good change. Now users have to be explicit how they want to encode binary data as JSON. Bas

Reply

Sign in to reply online Use email software

Conrad Parker

8:06 p.m.

On 23 September 2013 22:56, Bas van Dijk wrote:

...
On 18/09/13 04:02, Dan Burton wrote:

...
Everybody already uses, trusts, and loves aeson.

There's only one thing about it that I don't trust:

"Partial instances" that will call error in pure code and crash my

On 18 September 2013 06:30, Niklas Hambüchen wrote: program.

...
ByteString has such an instance that simply assumes that ByteString is the same as Text, so it calls Data.Text.Encoding.decodeUtf8 on it.

https://github.com/bos/aeson/issues/126 _______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries

Note that Bryan pushed a patch which removes the ByteString instances:

https://github.com/bos/aeson/commit/9ee73f303cacb8ae73b2325f2f47288522607420

I think this is a good change. Now users have to be explicit how they want to encode binary data as JSON.

great, +1 to including the next (major-bumped) version. Conrad.

Reply

Sign in to reply online Use email software

Yitzchak Gale

2 Oct 2 Oct

5:52 a.m.

Note that Bryan pushed a patch which removes the ByteString instances

Following Michael: +1 for including aeson in the next HP, even if due to dependency constraints that patch does not make it in this time. (But let's give it our best try.) -Yitz

Reply

Sign in to reply online Use email software

4248

Age (days ago)

4263

Last active (days ago)

Download

20 comments

17 participants

tags

participants (17)

Alejandro Serrano Mena
amindfv＠gmail.com
Bardur Arantsson
Bas van Dijk
Christopher Done
Conrad Parker
Dan Burton
David Luposchainsky
Edward Kmett
Herbert Valerio Riedel
Johan Tibell
John Lato
Michael Snoyman
Niklas Hambüchen
Simon Hengel
Thomas Schilling
Yitzchak Gale