Describe "RSpec's documentation" do

Testing RSpec
1

Hi all,

describe “RSpec’s documentation” do
it “should be helpful”
it “should be maintainable”
end

I’ve been wanting to improve RSpec’s documentation situation for a
long time, but my writing energies have been consumed by rspec itself
and The RSpec Book for a longer time, and will continue to do so for
the foreseeable future. So I’m going to need some help.

The problem to solve is that we have (at least) four sources of
documentation, each of which moves at its own pace and has evolved in
its meaning/place:

  1. The RSpec code examples that ship with RSpec
  2. The Cucumber features that ship with RSpec
  3. The github wiki: http://wiki.github.com/dchelimsky/rspec
  4. http://rspec.info

The model that Aslak and the Cucumber community has used has worked
very well because it’s a community effort, but there is still some
duplication between what’s on the wiki [1] and the Cucumber features/
scenarios that ship with Cucumber [2].

In the long run, what I’d like is the following:

  • Cucumber features that ship with RSpec become the authoritative end-
    user documentation. This is something that anybody can contribute to
    with patches, as it’s all in files that ship with RSpec. I’d also like
    to use such an effort to push the envelope on Cucumber features as
    executable documentation. I think that with a little bit of work we
    could use the features to generate a website with meaningful
    organization/navigation. Is anybody already doing that?
  • RSpec code examples become a solid source of additional detailed
    documentation for those who want to either extend RSpec or debug
    problems.
  • http://rspec.info becomes a one pager like http://cukes.info
  • The github wiki becomes a community driven resource center with
    links to tutorials, blogs, matcher libraries, etc, etc

I welcome suggestions, but I really need volunteers volunteers! I’m
not going to be able to spend much personal time on this, so if there
are any among you who are willing to coordinate with me and drive the
effort, I’d love to hear from you.

Cheers,
David

[1] http://wiki.github.com/aslakhellesoy/cucumber
[2] http://github.com/aslakhellesoy/cucumber/tree/master/features/

2

David C. escreveu:

the foreseeable future. So I’m going to need some help.
The model that Aslak and the Cucumber community has used has worked
as executable documentation. I think that with a little bit of work we
not going to be able to spend much personal time on this, so if there
are any among you who are willing to coordinate with me and drive the
effort, I’d love to hear from you.

Hi David, it is awesome that you are concerned about improving Rspec
documentation.

Although I have no time currently to coordinate/drive the effort, and
I’m not even skilled enough yet, I would be happy to contribute to
documentation when I get some time, slowly…

I really liked the docrails project and it was really simple to
contribute to Rails documentation using their model. Maybe it could be
another way of improving Rspec’s documentation. Although I like the idea
of using some kind of wiki, I simple don’t like the Github’s wiki… I
think they are a bit polluted…

It would be good if the Rspec site’s content were hosted on Github, and
there was a fork with open access for who wants to contribute, as it
happens on railsdoc, and then, from time to time, someone could take a
look at the changes and merge with main repository.

There are some mispelled words on Rspec site that could easily be
corrected that way… And I think we should mantain Rspec site,
improving its documentation instead of a one pager that links to
Github’s wiki…

Something like Rails Guides would also be awesome (like the tutorials
you’ve proposed).

What do you think?

Best Regards,

Rodrigo.

Faça ligações para outros computadores com o novo Yahoo! Messenger

3

This sounds pretty cool. Are there any ideas or examples of how features
would get turned into end user documentation? My first thought is
something
RDoc like would go in the free-form user story area… I’d certainly
like to
help as time permitted. I’d definitely like to be more informed of rspec
internals.

4

On Fri, Nov 6, 2009 at 5:27 PM, Ryan Briones
[email protected]wrote:

This sounds pretty cool. Are there any ideas or examples of how features
would get turned into end user documentation? My first thought is something
RDoc like would go in the free-form user story area… I’d certainly like to
help as time permitted. I’d definitely like to be more informed of rspec
internals.

Yes, we should definitely beef up the narratives in each feature. Then
we
need to generate the output as HTML and, the tricky part, generate some
wrapper HTML to make a site out of the different pages.

As for internals, my goal for the Cucumber features is to provide an
executable set of documentation for end users to understand how to use
RSpec. Not so much to expose internals, which I think is better
addressed in
the RSpec code examples themselves. Make sense?

5

On Sat, Nov 7, 2009 at 9:31 PM, David C. [email protected]
wrote:

As for internals, my goal for the Cucumber features is to provide an
executable set of documentation for end users to understand how to use
RSpec. Not so much to expose internals, which I think is better addressed in
the RSpec code examples themselves. Make sense?

I’m sorry David; it might be because I haven’t seen examples, but
conceptually I don’t think this quite gels. Some issues I have with
the idea:

1.) It’s trying to achieve too many goals at once. Application
verification and documentation are not the same problem. If you try
to do both with the exact same files, I believe both will suffer.

2.) It creates an unnecessary knowledge dependency. People who are
brand new to RSpec cannot be expected to have proficiency with
Cucumber just to read the docs. I understand that they don’t have
to execute the features, but if you don’t know Cucumber at all, trying
to wrap your head around the feature syntax will distract from the
part that’s important (understanding RSpec).

3.) The knowledge transfer would be inefficient, unclear, and
incomplete. Too many concepts cannot be effectively communicated in a
“Given/When/Then” structure. How do you explain the history and
philosophy of BDD in that syntax? How do you explain the role of
“red, green, refactor” patterns within iterative RSpec development?
–Presuming you’ve figured out how to phrase that knowledge in a
Cucumber feature, how do you execute it? And at that point what
have you gained?

Cucumber is great for many things. I don’t believe it would be a
great documentation tool at a “primer” level. If anyone can show that
it can not only be done, but be done well in a way that will make
inherent sense to newbies and speed up their learning of RSpec, I’d
stand corrected and support this wholeheartedly. Otherwise, I’d
suggest separating concerns, and not try to integrate and educate at
the same time.

—Having said that, I’m interested in contributing to the (prose, not
Cucumber) documentation effort.

–
Have Fun,
Steve E. ([email protected])
ESCAPE POD - The Science Fiction Podcast Magazine
http://www.escapepod.org

6

David C. escreveu:

…

As for internals, my goal for the Cucumber features is to provide an
executable set of documentation for end users to understand how to use
RSpec. Not so much to expose internals, which I think is better
addressed in the RSpec code examples themselves. Make sense?

Actually, I don’t like very much the idea of hiding the internals. I’ve
spent too much time trying to figure out what happens in the internals
so that I could integrate Rspec with Webrat and I still don’t understand
a lot of things.

For instance, why can I use ‘visit’ when I am testing an expected
behavior, but cannot use ‘redirected_to’ while both belongs to
Webrat::Session?

I think it helps a lot understanding what is happening on the internals.
Some topics that could be approached on documentation:

  • What happens on a describe/context/it/specify call?
  • What is the load order when running tests? (talking more deeply about
    spec_helper.rb)
  • How are database transactions dealt with in Rspec and the differences
    between before(:all) and before(:each) with respect to databases.
  • What is the class (how is it created?) that we are writing code on?
    How are the expectations/matchers injected on classes and which classes?

I really think that knowing the internals would help a lot and save us
many many time.

The syntax of Rspec should be simple so that reading the specs should be
natural, but it doesn’t mean that we should abstract from what is
happening in the internals…

Well, that is my opinion.

Regards,

Rodrigo.

Faça ligações para outros computadores com o novo Yahoo! Messenger

7

On Sun, Nov 8, 2009 at 5:55 AM, Rodrigo Rosenfeld R. <
[email protected]> wrote:

David C. escreveu:

As for internals, my goal for the Cucumber features is to provide an
executable set of documentation for end users to understand how to use
RSpec. Not so much to expose internals, which I think is better addressed in
the RSpec code examples themselves. Make sense?

Actually, I don’t like very much the idea of hiding the internals.

I’m not saying hide them. I’m saying document them in a different way.

  • What happens on a describe/context/it/specify call?
    The syntax of Rspec should be simple so that reading the specs should be
    natural, but it doesn’t mean that we should abstract from what is happening
    in the internals…

Separate, not abstract.

Well, that is my opinion.

Thanks for sharing it. I think we can work towards this goal as well,
but
I’d like to push this one further back as plan to make significant
changes
to internals in the next major version of rspec. We’ll roll that out
gradually (with alpha, beta, and candidate releases), but any time spent
on
a big document-the-internals effort before then would be a waste.

Cheers,
David

8

On Sat, Nov 7, 2009 at 11:39 PM, Stephen E. [email protected] wrote:

conceptually I don’t think this quite gels. Some issues I have with
the idea:

1.) It’s trying to achieve too many goals at once. Application
verification and documentation are not the same problem. If you try
to do both with the exact same files, I believe both will suffer.

Executable documentation is what I’m after. If Cucumber is not giving
that
to us now, then I’d like to use this opportunity to push in that
direction.
The whole point is to bind the documentation to the code so that as the
code
evolves, the documentation is less likely to stray from it.

2.) It creates an unnecessary knowledge dependency. People who are
brand new to RSpec cannot be expected to have proficiency with
Cucumber just to read the docs. I understand that they don’t have
to execute the features, but if you don’t know Cucumber at all, trying
to wrap your head around the feature syntax will distract from the
part that’s important (understanding RSpec).

Have you looked at
rspec/features/before_and_after_blocks/before_and_after_blocks.feature at master · dchelimsky/rspec · GitHub,
for example? I don’t think Given/When/Then are going to confuse anybody
in
this case.

3.) The knowledge transfer would be inefficient, unclear, and
incomplete. Too many concepts cannot be effectively communicated in a
“Given/When/Then” structure. How do you explain the history and
philosophy of BDD in that syntax? How do you explain the role of
“red, green, refactor” patterns within iterative RSpec development?
–Presuming you’ve figured out how to phrase that knowledge in a
Cucumber feature, how do you execute it? And at that point what
have you gained?

That’s what the narrative area is for, in part. And where that is
inappropriate, we should have a way to build a website that integrates
cucumber pages with plain HTML pages. Something like webby with a plugin
for
running the .feature files and storing the output as HTML pages. The
examples of what the code looks like should be executable.

Great. My goal is to have an integrated result, but that doesn’t mean
that
people can’t contribute to prose separately from scenarios.

–

Have Fun,

I usually do :slight_smile:

Cheers,
David

9

On 6 Nov 2009, at 12:49, David C. wrote:

In the long run, what I’d like is the following:

  • Cucumber features that ship with RSpec become the authoritative
    end-user documentation. This is something that anybody can
    contribute to with patches, as it’s all in files that ship with
    RSpec. I’d also like to use such an effort to push the envelope on
    Cucumber features as executable documentation. I think that with a
    little bit of work we could use the features to generate a website
    with meaningful organization/navigation. Is anybody already doing
    that?

+1

I had a little epiphany working on the wire protocol feature for
cucumber[3]. We were trying to design the protocol via email and
lighthouse ticket discussion, started using a wiki to document the
desired protocol when I realised - why don’t we just use the
features!? I would like to see more people pushing Cucumber in this
direction, and I think it would be interesting to consider adding mark-
up to comments to allow us to build RDoc-style websites from a
features folder.

drive the effort, I’d love to hear from you.

Cheers,
David

[1] http://wiki.github.com/aslakhellesoy/cucumber
[2] http://github.com/aslakhellesoy/cucumber/tree/master/features/

[3]http://github.com/aslakhellesoy/cucumber/blob/master/features/wire_protocol.feature

I’m well behind this effort and would like to offer my help on the
Cucumber end to make it possible to use the features to generate the
user documentation - I think this would make a terrific use case for
us to support.

cheers,
Matt

+447974 430184

10

On 9 Nov 2009, at 22:03, Matt W. wrote:

On 6 Nov 2009, at 12:49, David C. wrote:

In the long run, what I’d like is the following:

  • Cucumber features that ship with RSpec become the authoritative end-user documentation. This is something that anybody can contribute to with patches, as it’s all in files that ship with RSpec. I’d also like to use such an effort to push the envelope on Cucumber features as executable documentation. I think that with a little bit of work we could use the features to generate a website with meaningful organization/navigation. Is anybody already doing that?

+1

I had a little epiphany working on the wire protocol feature for cucumber[3]. We were trying to design the protocol via email and lighthouse ticket discussion, started using a wiki to document the desired protocol when I realised - why don’t we just use the features!? I would like to see more people pushing Cucumber in this direction, and I think it would be interesting to consider adding mark-up to comments to allow us to build RDoc-style websites from a features folder.

I’ve been looking at the documentation again recently, and I’d be happy
to start work on this, particularly if Matt W is wanting to look at the
cucumber end since he’s just up the road from me.

I’m very keen on the executable documentation aspects of this, and I did
a whole load of work back in 2008 on this kind of stuff using RSpec, so
I’ve got a bit of history here :slight_smile:

Matt

–
Matt P. | Design & Code
| http://www.reprocessed.org/

11

On 7 Jan 2010, at 00:53, Matt W. wrote:

So what’s the first step? A pint or two at the Reliance perhaps Mr Patterson?
That sounds about right :slight_smile:

Matt

–
Matt P. | Design & Code
[email protected] | http://www.reprocessed.org/

12

On 5 Jan 2010, at 12:17, Matt P. wrote:

RSpec. I’d also like to use such an effort to push the envelope on
desired protocol when I realised - why don’t we just use the
did a whole load of work back in 2008 on this kind of stuff using
RSpec, so I’ve got a bit of history here :slight_smile:

Great. I think it’s time we really started dogfooding cucumber in
terms of producing executable documentation. The features for Cucumber
itself could do with some housekeeping, but right now the focus there
is on the code, so RSpec seems like a nice place to work on this from.

I would like to see us try to build a flat HTML website (or PDF
reference book) from rspec’s features directory which could be used as
reference by a new or curious user. As David says, I think we could
drive out a bunch of really nice features from Cucumber to help make
it much easier to build and maintain large suites of features.

So what’s the first step? A pint or two at the Reliance perhaps Mr
Patterson?

cheers,
Matt

+447974 430184