State of UI Tests

[Stefano Mazzocchi]

Hi guys,

I tried to run the UI tests today and noticed that they are all broken. I'm not surprised since a lot has changed in the Refine UI and they were only marginally covering the UI anyway, but now we need to decide whether it's useful to have them or not.

In order to try the tests run

 ./refine ui_test

and it should install windmill and run stuff for you. There are 7 tests there, written in javascript with a wrapper around windmill that I wrote, the code is here

 http://code.google.com/p/google-refine/source/browse/trunk/main/tests/client/src/facets.js

Now, the question I have is: do we care about these tests? do we want to maintain them or expand them? or we should just avoid and kill the whole thing?

But at least we should be doing something about the fact that they are broken before the next big release.

Thoughts?

--
Stefano Mazzocchi  <stef...@google.com>
Software Engineer, Google Inc.

[Thad Guidry]

Kill them.  I do a LOT of testing of the UI, including internationalization and kick those issues to David all the time.  As long as I'm alive, I do not think the UI tests are that relevant.

--
-Thad
http://www.freebase.com/view/en/thad_guidry

[Tom Morris]

I don't know if Windmill is the right tool (not really my area of
expertise at all), but I think there is value in having at least some
level of end-to-end testing, so I'm loathe to just let the UI tests go
without a replacement. I just fixed a problem with parsing of form
parameters in multipart/form-data uploads which was preventing the
uploaded file from being named correctly. Perhaps someone who's more
familiar with Windmill can adjust the tests for the new UI.

As long as we're talking about testing, another whole that we
currently have is testing for the REST API between the server and the
client. We've kind of backed into allowing people to use this for
things like the Python client, which really means that we should be
testing it.

Tom

[David Huynh]

On Fri, Oct 7, 2011 at 4:36 PM, Tom Morris <tfmo...@gmail.com> wrote:

I don't know if Windmill is the right tool (not really my area of
expertise at all), but I think there is value in having at least some
level of end-to-end testing, so I'm loathe to just let the UI tests go
without a replacement.  I just fixed a problem with parsing of form
parameters in multipart/form-data uploads which was preventing the
uploaded file from being named correctly.  Perhaps someone who's more
familiar with Windmill can adjust the tests for the new UI.

I'm not sure if I even believe in automated UI testing :)

As long as we're talking about testing, another whole that we
currently have is testing for the REST API between the server and the
client.  We've kind of backed into allowing people to use this for
things like the Python client, which really means that we should be
testing it.

And extension developers also use these entry points ... I did try to make sure that projects can still be created by a single POST even if there's now an interactive UI. How about after 2.5, we document a subset of these entry points and have tests for them? We can grow the subset gradually.

David