I'm coming to the end of a large translation of 3 docx fies. I am unable to export the completed translation.
There's no error message, CT simply fails to create the exported document without explanation. (An error message would be nice.)
I've tried various target locations, I've tried export all and export individual document (I started out with the latter, as the customer wanted one document more quickly.)
Thank you for your detailed account.
a) CafeTran offers various rules: Sentence, Paragraph, Tag, Word, Document, Rules.srx, the default being Sentence. I can’t study the built-in, non-SRX rules, but if Sentence is what you use, it should not fail so much. I think it should at least split at periods, maybe unless there’s a non-breaking space involved?
For more information and suggestions on Segmentation, you can check: https://github.com/idimitriadis0/TheCafeTranFiles/wiki/1-Preferences#segmentation
b) For the demo version, yes. For the licensed version, you can use two or more concurrent installations with a ProZ Plus connection, but not with the classic offline license.
If you have such a license, you can test if anything breaks with the update by backing up the project folders (if outside the CafeTran folder) and copying the CTE installation folder or app, then launching the CafeTran executable from the copy and updating it.
If the ongoing projects export correctly, I think you can safely update the original installation. Again, keeping a backup around means you can always revert to it need be.
For the specific issue you describe, you may need to open a ticket for the developer to study the particular problematic project.
I posted an update soon after my earlier post, but apparently I forgot to submit it. Sorry for that. Here is the update to the update, also addressing your questions and suggestions:
1) I didn't update CT in the middle of the project
2) I never opened the source files created by CTE (much less changed their names). I did open the original files in my home directory, but this is not relevant, right? I didn't modify them, anyway.
3) As translation progressed, I exported intermediate versions of the files in the project, so that on Monday, at the time of the final export, I had some previous version of all of them (6 in total).
4) I then realised that the export did nothing, neither with the "export all documents" nor with the "export current document".
5) I tried quitting CTE and restarting it again: nothing happened.
6) I opened a new identical project and loaded the TMX, and decided not to pursue this solution as I was left with one third of the project untranslated. The reason is manual split and merges: when I create the project, some segments are very large, consisting of several complete sentences, and other segments consist of split sentences. I haven't looked into this yet as I have always used the default segmentation rules; I will certainly have a look at your suggestions on this point.
7) At some point (maybe after some other attempts - I'm not sure) I glued, then unglued one by one the documents and tried again single file exports: this worked for 5 files out of 6.
8) For the sixth file, I loaded a copy of it and manually altered the segments (same I should have done in point 6) above, but on a much shorter scale).
So that's it. The answering your questions:
- I didn't update CT
- I didn't move or touch the source files
- I use the default segmentation rules and manually change it when necessary
- I didn't try a computer restart, only a CT restart.
- I tried the QA check when attempting to create a new project, but it didn't solve much as the problem clearly lied elsewhere (segmentation)
- My Java memory is 2048
- Something changed when gluing/ungluing.
Thank you all for you help. I'm left with two questions:
a) Any obvious reason why segmentation should fail so much?
b) I know that one should not update CafeTran in the midst of a project. How do you recommend to proceed if that's not possible? Say, if one works on large projects spanning weeks or months and partially overlapping. Is it possible to simultaneously use two different versions of CT?
> This information should be provided in an error message, not on the forums.
I agree. That error handling might be included in a future update.
> You often suggest this as a posited source of errors for all sorts of issues.
I don't remember suggesting it as a source of the error other than with the Export.
Igor: "The final export may fail due the following (and known) reasons:"
This information should be provided in an error message, not on the forums.
Most of the above issues are fixable if the user were to be provided with an appropriate error message.
Re. 1: "1. Modifying or renaming the source language document located in the project folder in the middle of the translation - before the final export. The obvious solution is the replace the current version with the changed source document in the project via the Project > Replace document.. ."
You often suggest this as a posited source of errors for all sorts of issues, but I find it unimaginable that anyone would ever do this. Has anyone ever actually done this?
> she clearly stated, she hasn't modified the original source language document...
As it is not so clear to me, I assumed one of the source language documents may have been changed somehow (or perhaps the program was updated in the meantime?).
> you don't address her actual problem – that the export function in CT sometimes fails for unknown reasons – at all.
The final export may fail due the following (and known) reasons:
1. Modifying or renaming the source language document located in the project folder in the middle of the translation - before the final export. The obvious solution is the replace the current version with the changed source document in the project via the Project > Replace document.. .
2. Changing the names of the project files or paths (e.g. when moving projects across computers or operating systems, clouds etc).
3. Not enough Java memory assigned for the program by the user (e.g for larger projects) in Preferences > Memory tab > Java memory field.
4. The user updated CafeTran from an older version. It is always advisable to complete the projects before updating.
A tip: for larger projects, it is good to do the export and check it periodically (from time to time - especially when updating the operating system or the program).
A few questions to boot:
- Did you update CafeTran while working on this project? Users are advised to complete all their translation projects in their current CafeTran version before updating. Case in point: There has been a recent change in Word and LibreOffice file filters.
- Did you move or edit the source files contained in the CafeTran folder for this project? These files should be left untouched.
- Did you run a QA for your project? Especially, make sure you run QA > Tags check (and eventually At tags spaces check).
- What segmentation settings have you used and what changed have been conducted? (Manual split and merges?)
- Finally, just re-try after a computer restart.
sorry to butt in, but you've completely misunderstood valentina's post.
1. As she clearly stated, she hasn't modified the original source language document, she's modified the segmentation in CafeTran.
2. She doesn't "wish to re-create the project with the TMX memory", she has been forced to try this workaround because your software doesn't work as it should. She correctly anticipates problems because she has extensively modified the segmentation in CT.
3. Other than by offering an extremely time-consuming workaround (essentially telling her to do it manually) you don't address her actual problem – that the export function in CT sometimes fails for unknown reasons – at all.
P.S. @Valentina: a lot of segmentation problems can avoided by updating the segmentation rules file – see this post, in particularly the second comment by Jean Dimitriadis
> I've largely modified the segmentation with respect to the original files
This is not a good idea to change the segmentation if you wish to re-create the project with the TMX memory. The export can fail if you modify the original source language document located in the project folder in the middle of the translation (before the final export). I would recommend loading the older project and trying to export each file individually - the menu Project > Export and exchange > Export current document (after unglueing the single document via the Project > Documents menu). If that doesn't help, you can try to export to HTML or Text format. Then, open the exported (HTML or Text) document in Word, copy the target side and paste it into the new Word document. Finally, review and adjust the formatting if needed.
I'm having the same issue with a large project consisting of several Word files, which I've just finished. The export command simply ignores me. I've tried to bypass the problem by opening a new, identical project and using the original TMX to translate it, but it won't work since I've largely modified the segmentation with respect to the original files. Any suggestions? Thanks.
> I assume it's just a quesiton of using a relative, rather than absolute, path?
Yes. With the glued (multiple) documents, CTE looks for the absolute path. If you translate a single document project, the program can check both - first absolute, then relative if the absolute one is not found. I plan to look into the fallback to the relative paths for the glued mode as well.
> Did you edit the project in another translation tool?
> Did you change anything (contents, name or location) of the original (source language) documents in the middle of their translation?
> Did you move the project files across the computers or operating systems?
Yes, I synced between my two computers. And this must be problem, as I've just got home and tried the export on my desktop at home and it works fine.
It would be good if there was an error message for this
problem. On doing the export on this machine, I now realise that the failed export on the other machine
didn't display the 'coffee time' message when the export failed. An
actual error message, ideally with the above information, would be good.
In all other respects, syncing between computers works fine (except for having to reselect the memory and glossaries (?) when switching betwen machines.
Would be good if this worked too. (I assume it's just a quesiton of using a relative, rather than absolute, path?)
Did you edit the project in another translation tool?
Did you change anything (contents, name or location) of the original (source language) documents in the middle of their translation?
Did you move the project files across the computers or operating systems?