Hi ,

I have tried to find others who may have raised this question before- or indeed mentioned performance issues, but have come up blank on this one. Therefore, I suspect any performance issues would be caused by the file(s) itself and not through the virtual merge functionality.

It would be good to hear from real life translators with experience of virtually merging large numbers of files to understand their experiences.

Hi,

We always merge files (recently with a total wordcount of > 60,000) and cannot confirm that they behave differently from individual files. The benefits of merging are obvious (AutoPropagate and filtering) so I can only recommend it. On our machine, we sometimes experience display problems at the end of a large merged file (> 60,000 words), but these may be related to insufficient RAM.

On the other hand, I recently opened a set of merged files and then another smaller file in a separate tab. When changing between the tabs, I always had to wait a while (probably the data is loaded/unloaded in the background).

HTH

Best

Annette (from Manfred's account)

Hello Manfred Altmann

I have a sample case with about 650 mostly very small XML files, some with embedded HTML (using the embedded content processor). Switching from one segment to the next takes more than 10 minutes. Not workable.

The same content in one xml file is okay to work with. (About 30K words. Neither CPU nor RAM are maxed out.)

Strange, is it not?

Daniel

Hello Manfred Altmann

I have a sample case with about 650 mostly very small XML files, some with embedded HTML (using the embedded content processor). Switching from one segment to the next takes more than 10 minutes. Not workable.

The same content in one xml file is okay to work with. (About 30K words. Neither CPU nor RAM are maxed out.)

Strange, is it not?

Daniel

Another test: subset of 30 XMLs virtually merged with a total of 892 words. Time for switching from one empty segment to another is 2 minutes. I tried this both with and without embedded content processor - it does not make a noticeable difference.

Paul Filkin I am looking for performance related defects on JIRA. Translators are complaining about translating hundreds of small XML files in a job with large total word count, meaning a mojority of the content is 100% matched and word count for the job is small. So they are spending more time just managing file size, rather than translating. It's not an attractive situation for translators. Can you help?

Thanks,

Naoko

Naoko Jones

There was one for Studio 2017... CRQ-10687.  I can't find any others that are performance related. Why don't you log a support case and then you'll find out forsure as the support teamwill investigate more thoroughly.

Paul Filkin

Thanks very much for your quick reply! This is exactly what we needed, our program manager can add her comments to it and we can go from there.

Thanks,

Naoko

Naoko Jones, Paul Filkin

Would this ticket address my issue as well? I am still experiencing crippling performance issues with many virtually merged files (xml or icml usually). There seems to be a threshold where the performance drops quite suddenly, maybe around 20 merged files.

I also have the suspicion of a memory leak or something like it, as the problem worsens over time and is better once I restart Studio (but then worsens very fast).

I tried storing files on this drive or that - to little or no effect.

Daniel

Unknown said:

Would this ticket address my issue as well?

I'm doubtful as this ticket is a 2017 issue and there is a note that specifically says the problem didn't occur in 2019.  I think, if you're experiencing prolems like this, that you should raise a support ticket for this.

Paul Filkin

I can't submit tickets because I don't have a support contract.

Daniel

That's unfortunate!  If you can send me the project and steps to reproduce I can do this on your behalf to make sure we capture this problem and ensure it's on the dev teams radar.

Paul Filkin

I sent you some files.

Daniel

Thanks

I created a project and opened all 552 files for translation with a virtual merge.  I added an empty TM and then used LC NMT to start translating.  Once I'd translated around 150 segments things did indeed start to slow down to a crawl.

Is it ok for me to share this project with the support team?