[reportlab-users] Automatic testing PDF output

[Stephan Richter]

On Friday 22 May 2009, Andy Robinson wrote:

> Just knowing that 'something changed' in the full test suite output of

> rst2pdf, z3c.rml or even our own test suite is of limited use.  There

> are a ton of spurious changes at the pixel level when you update

> libraries involved in the rendering of the text as well as the

> generation (or switch fonts at someone else's behest!).   We've tried.


Me too and I am claiming it worked quiet well.The picture comparison worked on 
multiple OSes (including Wondows) over various OS upgrades in Linux.


> What we could really use is a collection of small scripts or

> functions, each exercising one carefully-thought-out "feature that

> matters", with a brief note underneath saying what the inspector

> should see.   For example, the table I posted yesterday and the

> statement 'all cells in this table should have the same baseline'.


Well, in my case the RML documents that. Since I usually demonstrate a single 
feature, my expected output RML and PDF are pretty simple too.


> We can then link these up on our servers to show a gallery of images,

> tied together with HTML, allowing easy visual inspection and also

> drawing the eye to differences between (say) trunk and last release.

> And when there is a change, a human can sort through fairly easily

> deciding if they matter.


yep. I really liked Marius' diffing technique. I would love to integrate that 
in z3c.rml when a failure occurs.

Regards,
Stephan
-- 
Entrepreneur and Software Geek
Google me. "Zope Stephan Richter"