parseGPL reparses GPL files in the same session about geoquery HOT 6 CLOSED

My suggestion to optimize this is to have an internal environment where the parseGPL (or maybe the .parseGPLWithLimits) function could check for a cached version of the GPL object read from the file path passed as argument.

Some possible details necessary for this feature to work are:

• distinguish between objects that were parsed with different number of rows limits
• give the user an option to not use the cached version
• even if not explicitly specified, ignore the cached version if the original file was modified

Do you think this is reliable? It's just an initial idea...

from geoquery.

There are really two issues you have touched on:

1. Avoiding parsing the same GPL multiple times
2. A caching mechanism that is smart enough to keep GPLs around throughout a session.

I like the idea of number 1, but I would be concerned that keeping GPLs around in memory (number 2) might end up biting some users who don't realize this in-memory caching is going on. I think what might make sense is the ability to pass in a list of GPL objects to getGEO() and getGEO would then check if a required GPL is already in the list. If so, getGEO would use it and not download and reparse. If not, the GPL would be fetched. In a sense, we would leave caching to the user, making it less magical and less likely to catch new and "casual" users off-guard.

So, I would propose adding another argument to getGEO:

getGEO(..., GPLlist = list(GPL, GPL, GPL, ...))

What do you think, @leogama?

from geoquery.

from geoquery.

from geoquery.

Sorry to take so long to get back to you. How would you want to do the caching? In memory? On disk? If on disk, in what format?

from geoquery.

I think @seandavi and me forgot to close this issue after merging #24. Mr. @assaron even developed his own file cache solution.

At the time, I thought the message displayed at the package loading was enough to give users a clue about this functionality. Should it be explicitly documented somewhere? Or maybe just add the word cache to the option's name...

from geoquery.