Hi Owen,
On Thu Dec 18 12:13:44 2014, owenhollands@gmail.com wrote:
Show quoted text> As an asside, do you think there is value in a check script that
> compares the HTML off the bankwest website with a reference copy to
> try and detect of the website ahs been updated?
I suppose it would be a bit complicated to directly compare live website
data to a reference copy, because of the varying tokens, account info,
etc. found on every page.
On that basis, I would probably go about this by scripting a bunch of
actions on the Bankwest website, passing the resultant HTML to the
Finance::Bank::Bankwest::Parser modules, and then testing that they
don't reject the HTML.
That alone, though, probably wouldn't achieve anything more than just
regularly using the Finance::Bank::Bankwest distribution normally.
I've tried to code those parser modules with enough expectations that
they fail quite verbosely if Bankwest significantly alters the
construction of its responses, but with enough leniency that minor
design changes would allow the code to continue to work. It seems to be
a difficult balance to get right.
The parsers continued to work in this case because the change Bankwest
made was extremely subtle: take a <select> box which was previously
optional, add a new <option> and require that <option> to be selected
for the date fields not to be ignored. As far as the relevant parser
was concerned, it was exactly the same page.
In short, I'm not against the idea of some sort of check script, but I'm
struggling to picture how a useful one would operate and whether it
would really result in fewer breakages. I'm definitely open to further
discussion and insight on the matter though. Thanks for the idea; I'll
keep thinking on it.
--
Kind regards,
Alex Peters
