Re: Bug 85/4494 (keeping track of validation statistics for various purposes

Hi Brian,

On Feb 7, 2008, at 02:17 , Brian Wilson wrote:
>
> I wanted to do this too, early on. I tried to customize the templating
> system myself and intercept the arguments that were being passed via
> error_messages.cfg, but I just did not understand the way things were
> working. Specifically, in
> http://dev.w3.org/cvsweb/validator/share/templates/en_US/error_messages.cfg?rev=1.32&content-type=text/x-cvsweb-markup
> I wanted to preserve all the %1, %2, ... arguments (It looks like  
> err #136
> has the most arguments at 6). While it seems esoteric and totally
> pointless to probably *everyone* else's needs, adding some sort of
> abbreviated message of this type to SOAP might be interesting:
>
>            <m:error>
>                <m:line>596</m:line>
>                <m:col>1169</m:col>
>                <m:message>end tag for &quot;UL&quot; which is not
> finished</m:message>
>                <m:messageid>73</m:messageid>
>                <m:messagearg>&quot;UL&quot;</m:messagearg>
>                <m:explanation>[stuff deleted]</m:explanation>
>                <m:source>[stuff deleted]</m:source>
>            </m:error>
>
> where each successive m:messagearg element captures the variable  
> arguments
> used in the error message. This is much more compact than storing the
> entire error message.
>
> Ignoring (for the moment) whether such a feature addition would be
> useful to anyone else's needs, would that be hard to do? If that  
> could be
> added, I could grab that information in a future crawl.

Short answer: yes, it would be a bit hard.
Long answer: it's in my todo list.

If you looked at share/templates/en_US/error_messages.cfg you may have  
been thinking that the %1, %2 are being used by the validator to   
create the error messages. That's actually not the case yet. The  
validator does use the error explanations from that file at the  
moment, but the error message itself comes directly from the various  
parsers. So the validator doesn't actually "know" that the mesage is  
about UL, it just passes the error message.

For now.

However, for the sake of e.g localization, we'll need to make the  
validator more aware of the text and context of the error messages,  
which is why it's on my to-do list.

Hope this is clear(ish).
-- 
olivier

Received on Monday, 18 February 2008 04:59:48 UTC