Re: [tap] RFC Status?

[Steffen Schwigon <[e-mail]>]

Leon Timmermans <[e-mail]> writes:
> On Fri, Sep 19, 2014 at 5:59 PM, Steffen Schwigon <[e-mail]> wrote:
>     And here I seriously start to panic if that opinion should ever lead to
>     deprecating YAMLish, as 1) it much better fits the philosophy of TAP's
>     line-based'ness, 2) is easier to generate with line based tools, like in
>     a shell-only world, and 3) zillions of test results are already based on
>     it for the last 6 years. I'm not exaggerating here.
>
> I wasn't aware of anyone actually outputting YAML. Can you tell us more about this?

With great pleasure. :-) Here we go:

The Tapper test infrastructure (http://tapper-testing.org/) consumes it
with the help of TAP::DOM, Data::DPath (and a handful marines), and it's
generated for embedded benchmark data in projects like Perl::Formance
(http://perlformance.net/), general supporting data in libs like
"bash-test-utils"
(https://github.com/tapper/Tapper-autoreport/blob/master/bash-test-utils.md)
and similar, but less public projects. In general we use it in a
restricted world like OS testing with only shell/printf/printk
available.

These projects might not look very active but that's only lack of github
pushes due to lack of tuits (usually just once a year during the Perl QA
hackathons). They are in very heavy use.

Slide decks (bit old but still valid):
 - http://tapper-testing.org/docs/yapc_eu_2011_tapjuggling.pdf - pg. 23, 37, 41
 - http://tapper-testing.org/docs/linuxcon_eu_2011_linux_and_virtualization_testing_with_tapper.pdf  pg. 35, 43

Two details make it harder to generate JSON than YAMLish:
 - on generating: JSON is oversensitive to trailing comma before closing brackets/braces.
   If you generate it from templates, like Template-Toolkit, you go mental.
 - on consuming: JSON defines Javascript-specific thingies, in particular true/false.
   Sounds minor, but makes it hard in restricted worlds like OS testing.

Kind regards,
Steffen
-- 
Steffen Schwigon <[e-mail]>