several tweaks, mainly saving method results for re-use, to improve #1752
Conversation
performance of climatology etc. on Rhea or other machines where I/O is expensive.
| # For isTime(), keep track of whether each id is for a time axis or not, for better performance. | ||
| # This dictionary is a class variable (not a member of any particular instance). | ||
| idtaxis = {} # id:type where type is 'T' for time, 'O' for other | ||
|
|
@painter1 what happens if you have mutliple files with same axis id, but only some are actually time? Wouldn't this break? I'm thinking auto generated arrays axis_0 etc...
Yes, it would break! I've never seen such a thing, but among the changes I've made, this is the one which I was least sure about.
maybe we should add a test? See if it's already in and if it is and type changes, throw a warning/error? I don't think it will ever happen but of course now that I said it it is bound to happen...
The point of this is to not have to figure out what the type is - that involves reading from a file. So the way I would fix it is to use self.idtaxis only if the user does something to enable this feature, e.g. calls a method to initialize it.
we are now caching this after first use
doutriaux1
added
3 - Review
1 - Ready
2 - Working <= 5
and removed
3 - Review
1 - Ready
labels
|
I will have a look at it for review. |
|
Is this good enough to be merged? We need this for release 2.4.1, and the code freeze is tomorrow. |
|
@painter1 it looks good to me, certain code I couldn't reasonably test since its very specific but overall its good. +2 |
|
Our Mac bots are hosed but I am okay if we merge it as it is. |
several tweaks, mainly saving method results for re-use, to improve
performance of climatology etc. on Rhea or other machines where I/O
is expensive.
WORK IN PROGRESS DO NOT MERGE YET
The text was updated successfully, but these errors were encountered: