Patches

[Richard Kimberly Heck]

On 11/6/20 11:13 AM, Richard Kimberly Heck wrote:
> On 11/5/20 5:43 AM, Yuriy Skalko wrote:
>>> Yes, updateMacros is a pain, it is actually a O(n^2) algorithm which 
>>> kills performance for big documents. But it is a bit scary and I 
>>> never dared changing it :)
>>>
>>> JMarc
>>
>> Yes, macros implementation is really tangled. Maybe at first we can 
>> try to minimize calling updateMacros?
> That is definitely the goal. But, as you saw with 9e7832915f, this is 
> all very fragile.
>
>> Also attaching several patches with some improvements.
>
> These all look ok to me, so go ahead with them, with the exception of 
> #2. I *think* this is probably redundant---it was done in 
> updateBuffer, yes---but other things were done after that, so it may 
> or may not actually be redundant in this case.
>
> But I'd suggest we not try to untangle this further until after 2.4.0 
> (or that all such work go into a feature branch). This is 
> super-dangerous stuff. But maybe if a bunch of us went at it shortly 
> after that release, we could make some progress.

Toward that end, I've created a features/cleanup/updateMacros branch and 
committed a very small piece. The first step, for me, is to figure out 
how this stuff works!

Riki