Performance drop-off between versions

[Heather Wallace]

Did something fundamentally change in the way Word documents are populated between version 3.3 and 3.4? I’ve been investigating performance using eaDocX between 2 different computers. The original was still using 3.3.15.3, but the faster computer used version 3.4.3. I decided to do some specific trials on the faster computer to eliminate version from the comparison, but also out of interest.

My trial document contained some boilerplate text, 6 package-driven sections including diagrams, child elements and tables of relationship elements, 11 element reports of varying sizes and a glossary. it comprises over 400 pages.

In version 3.13.3 this took only 11 minutes to generate
In version 3.4.3 this increased to 23 minutes
In version 3.5.0.3 this remained at 23 minutes.

A doubling of generation time was not what I expected from an upgrade. Are any measures being taken to ensure performance does not slow further?

Regards,
Heather

[eadocX Support]

To make this a fair test, can you run it on the same machine, with just different eaDocX versions?

[Heather Wallace]

That’s what I did. I installed each version in turn using our currently spare license on my home computer, where I have admin rights. Shortly that license will be used on another computer in the office, again with admin rights (I’ve uninstalled it at home). Do you want me to rerun the tests?

Heather

[eadocX Support]

Ah – I misunderstood the original post.
Short answer is that I have no data on expectations of performance between 3.3 and 3.4 – I’m working on 3.5, and that’s changed all over the place.
So no help from me, sorry.

[Heather Wallace]

For your information, I did a rerun on the new computer and paid closer attention to the time splits. The major increase in time relates to the “generate cross-references” part of the process.

Heather

[eadocX Support]

Ah – that’s interesting.
Creation of cross-references is where eaDocX is most sensitive, as each time it resolves a cross-ref, we have to make a call to MS Word, and these are expensive (a few 100ms each call), which is why we have the ‘Draft’ document option which lets you produce docs with not cross-references more quickly.
I can’t think of any specific changes to the cross-ref mechanism between the releases you mentioned, but it’s in the larger documents that we start to see different effects, as working arrays of variables get larger etc.
For the next non-maintenance release, I’ll put the performance analyser on the code, with a larger document, and see if there are any surprises.
Thanks

[Heather Wallace]

Cheers,

In the meantime I’ll try doing the whole document in draft mode to see if there are any issues (there are no explicit cross references, but I remember from early experiences that there can be an impact on table formats).

Heather

[Heather Wallace]

OK, that works well. Back down from 36 minuts to 13 using draft mode with no ill effects on the big docuemnt. Unfortunately some of our documents contain matrix reports. These are not formatted by the draft generation process. I guess it will be more time-effective to generate these in a dedicated run as formatting of just the matrices still occurs quickly.

Thanks again,

Heather

[Author]

Posts