Page 1 of 1
faster batch

Posted:
Wed Jan 18, 2012 11:09 pm
by roedygr
I have often asked that you cache results from a batch validation so that if you repeat without the file changing you can do it instantaneously.
Here are are three similar but weaker requests that would be easier to implement.
1. have an option for RECENT that validates anything that has changed since the last Batch validation run.
2. (1) PLUS validate any files that had errors on the last batch validation, even if they have not changed since.
3. lIke (2) but include ALL files that failed their validation, no matter how long ago.
With these, you don't need to cache results, just a time and possibly some file names.
Re: faster batch

Posted:
Wed Jan 18, 2012 11:41 pm
by roedygr
Here is how you would implement it.
You tack a timestamp and list of files onto the batch descriptor. Any time the user changes the criteria, you discard the timestamp and list.
The first time you run, it works as now. But when you are done you put the time the run started and the fully qualified filenames of the files with errors or warnings on the end of the batch file.
The next time you run, the files you check are the list of bad files, plus the list of files conforming to the batch parms that have changed since the last run. You validate just those files. You know the others will have to be good. When you are done, you add any files you found bad, and remove any you found good and refresh the list on the tail end of the batch file. If the user aborts part way through, to keep things simple, you can just leave the list of bad files and the date as is. You treat it as if the aborted run had never been done. If you want to get fancy, you can remove bad files found good as you go. You need to avoid dups and you need fully qualified filenames since the CWD might change between runs.
Re: faster batch

Posted:
Thu Jan 19, 2012 11:11 am
by Albert Wiersch
Hi Roedy,
Thanks. That gives me some ideas that might not be too difficult or time consuming to implement - like the automatic creation of another target list that contains only targets with errors or warnings.
I'll try to come back to this for v12 and try to implement something if time permits.
Re: faster batch

Posted:
Thu Jan 19, 2012 1:42 pm
by roedygr
There is another benefit to implementing this. As a side effect get a list of files with problems in a given batch. You could then with one click load them all into the editor for repair and close with ALT-W.
The scheme pretty much cuts out ever validating a file that is already known to be good. I would would bet Validator spends 95% of its time now revalidating files that could not possibly contain errors. The net effect should be equivalent to speeding up HTMLValidator by about a factor of 20.