Support forum of the software localization tool Sisulizer


.NET, Delphi, ... - Sisulizer Localization Tool Support Home

Get in contact with the makers of Sisulizer.
Our forum is open for all questions around Sisulizer from customers and prospects.
Don't hesitate to register and ask. The Sisulizer team will answer ASAP.

Search     Help Home Sisulizer Website Download
Search by username
Not logged in - Login | Register 

 Moderated by: Renate.Reinartz, Markus.Kreisel, Jaakko.Salmenius, Ilkka.Salmenius
New Topic Reply Printer Friendly
Help topics "changed" incorrectly after scanning of new help file - Bugs and Quirks in Sisulizer - Technical Support (You need to be registered at the forum to write) - .NET, Delphi, ... - Sisulizer Localization Tool Support
AuthorPost
 Posted: Tue Jun 22nd, 2010 10:31 am
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
Hi,

We scanned an existing project for changes, and in the resulting output, there are multiple examples where Sisulizer marked strings as changed (when they weren't) without including the existing translations in the "changed" strings. The strings with the existing translations are marked as unused instead. See the example in the image below. The "English" column shows how Sisulizer interpreted the strings in the previous scan, since the translated text there was simply copied from the previous "Original" column. English was added for the purpose of seeing which strings have actually changed.

I can understand that the two sentences should be addressed as separate strings/segments (that should have been done in the first go-around as well), but in this case, even though Sisulizer appears to have possibly perceived the strings as separate, it has only removed part of the string and then reproduced the entire string again. In Sisulizer, if you click the button to show the original while displaying the node in question, both strings are displayed even though they don't both occur in the real help file.

In addition there are many strings that are marked as changed that appear to be identical, but that will be (re)-addressed in a separate issue.

Allen

Attachment: Help_scanning_error.png (Downloaded 69 times)

Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 11:53 am
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
Hi Allen,

Sisulizer for specify location of string in source needs to know context of this string. This is not problem for binary Delphi, NET sources or .NET projects, because these platforms always use unique contexts for strings. However, HTML platform doesn't use unique context. So, Sisulizer needs use own methods for assigning unique context for HTML platform. Our developers continuously improve this functionality, but this is very hard job and in some cases (many changes in source file structure and changes in originals) Sisulizer still can lose translations.
Could you upload to us new, old (if you still have this old file) versions of your CHM file and SLP file? We try monitor similar cases and maybe your files can help us improve context functionality for platforms without real context.

Best regards,
Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 12:03 pm
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
One additional question:
Did you change recently segmentation settings  in your project?

Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 12:05 pm
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
Hi,

Not knowingly at least:?.

Allen

Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 12:31 pm
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
OK, thanks. I'll check all possible tracks with your project (of course, if you send it).

Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 12:53 pm
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
Hi,

I uploaded the files that I believe are closest to the ones that were used to get the results in question.

Best regards,

Allen

Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 01:16 pm
PM Private Upload Quote Reply
Markus.Kreisel
Administrator


Joined: Sat Apr 8th, 2006
Location: Monschau, Germany
Posts: 2143
Status: 
Offline
Hi,

In your file DOX64_help.slp is segmentation switched off.
In DOX63_help.slp it is switched on.

Check Project - Edit Source - <your source> - Segmentation.

Can you please verify that it is the same in your files.

If segmentation changes the segments do not fit anymore. Hope switching it on will do the trick and scan for changes finds your strings again.

I also saw that your HTML source slightly changed in formatting. Did you use some HTML code optimization tool? Did perhaps Robohelp has some new option?

Markus



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 22nd, 2010 02:36 pm
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
Hi Markus,

The DOX63_help.slp file was the wrong file. I just uploaded the correct one. Segmentation is turned in it as well.

Allen

Back To Top PM Private Upload Quote Reply

 Posted: Wed Jun 23rd, 2010 07:55 am
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
Hi,

To answer your question about the formatting changes you saw, nothing was done differently, at least, and no additional tools were applied, but having said that, it might be possible that a RoboHelp service pack was installed between help versions, but I don't know for sure that that is the case here.

Allen

Back To Top PM Private Upload Quote Reply

 Posted: Wed Jun 23rd, 2010 08:06 am
PM Private Upload Quote Reply
Markus.Kreisel
Administrator


Joined: Sat Apr 8th, 2006
Location: Monschau, Germany
Posts: 2143
Status: 
Offline
Hi,

In your file DOX64_help.slp is segmentation switched off.
In DOX63_help_updated.slp it is switched on.

Check Project - Edit Source - - Segmentation.

If segmentation is different then the rows in your files are different. That means that Sisulizer can not make matches by value (since it is document localization).

Please set segmentation to on before you do a scan for changes. I assumed that DOX63 are the old files and DOX64 is the new one.

Markus



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Wed Jun 23rd, 2010 11:22 am
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
I reproduced this issue with my own, sample project based on your CHM file, so issue is not related to potential changes in segmentation settings.

Best regards,
Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Wed Jun 23rd, 2010 12:11 pm
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
Hello Allen,

I found reason. Issue is result using of segmentation rules (not changed) for changed source file. Below is small example.

1. Source file contains following item:
<span>Lalala lalala.</span> Ble ble ble.
Because directly after dot is tag not space, Sisulizer don't use rule for breaking sentences in paragraph to separated segments.

2. After update of source file this item look like this:
<span>Lalala lalala. </span>Ble ble ble.
Sapce was moved before tag, and for this reason Sisulizer use default rule for breaking sentences in paragraph to separated segments.

However, I think Sisulizer should ignore tag items during checking with segmentation rules. I'll forward it our developers. I hope R&D will find solution for this issue.

Workaround:
1. Create custom exception rule for this case. In this case Sisulizer should revert your project to previous state. You can learn how create custom segmentation rule in this http://www.the-localization-tool.com/?p=624 article on our blog.
2. Disable using of segmentation rules in this project. This solution doesn't revert your missed translations (it can even generate next not translated items), but this solution allows you avoid similar problems in future.

Best regards,
Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Wed Jun 23rd, 2010 02:26 pm
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
OK, sorry. I misread your earlier mail, and you're right, segmentation is turned off in the 64 file, but after having looked at it closer, segmentation turned off appears to be the default setting which then makes sense why it was turned off, since we we obviously not very aware of it. The question then, I guess, is how it got turned on in the new file in the first place. Oh well... According to Janusz, that's apparently not the issue anyway:)

Here are some statistics that might interest you. I took the original 6.3 file (with segmentation turned off) and saved it as two new files - one with segmentation turned off and one with segmentation turned on. I then scanned each one for changes, and these were the results:

With segmentation turned on (this is apparently what we did before since the numbers were about the same):
  • About 915 new strings and 1443 strings marked as changed. Having gone through this already, we know that about 10% of the strings in the English "translation" are identical to the original strings, about 1100 of the strings have only been changed by a space being moved inside or outside of a SPAN tag (which we know we didn't do in the source files), and only about 200 strings actually changed.

    I wonder if Sisulizer is inadvertently changing the position of the spaces in the second scenario during scanning to make the strings different from the previous scan?
With segmentation turned off (i.e., the same setting as the original file):
  • More of the strings appear to match even though the problem of adding extra strings still remains. There are over a 1000 new strings and 2442 strings marked as changed. I would have expected the opposite to be true since the segmentation settings were the same in this case. This must mean that there are a considerable number of additional "changed" strings that are exactly the same as the original.
I don't know if that will be of any use to you, but there you have it.

Best regards,

Allen

Back To Top PM Private Upload Quote Reply

 Posted: Wed Jun 23rd, 2010 03:16 pm
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
Oh well... According to Janusz, that's apparently not the issue anyway
This is real reason of issue visible on your screenshot. My tests were concerned to these items (from screenshot).

I wonder if Sisulizer is inadvertently changing the position of the spaces in the second scenario during scanning to make the strings different from the previous scan?
No, differences are related to source files created by RoboHelp.

There are over a 1000 new strings and 2442 strings marked as changed. I would have expected the opposite to be true since the segmentation settings were the same in this case. This must mean that there are a considerable number of additional "changed" strings that are exactly the same as the original.
Both CHM files are different. Deferences are not only concerned to contents of existing HTML files, but also to count of included HTML files. On below screenshot you can see contents of this same directory decompiled from your new and old CHM files. Old directory contains 16 files, new contains 23 files.

Janusz

Attachment: screen.png (Downloaded 45 times)



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Tue Jun 29th, 2010 01:05 pm
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
Hello,

Segmentation issue was fixed in build 307. Now, if string (HTML) contains sub tags e.g. <b>...</b>, text won't be segmented. New build will be released soon.

Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Wed Jul 7th, 2010 03:09 pm
PM Private Upload Quote Reply
a_keann
Member
 

Joined: Wed Nov 19th, 2008
Location:  
Posts: 108
Status: 
Offline
Hi,

I think we may have finally found the root of all evil in this issue which has caused us literally weeks of work.

Since segmentation is turned off by default when a project is created (it could be good to provide the option to select the opposite during the creation process), segmentation in the original file in question must have been turned off. At some point, either we inadvertently found the setting to segment strings in HTML and turned it on, or perhaps the setting was changed somehow due to a previous build. I don't have any recollection of having done that, but either way, we ended up going through the file to fix all of the changes that resulted when the project was updated the first time (which may have been the reason for the first forum entry on the subject http://www.sisulizer.net/forum3/1848.html). 

However, during the process of working with this, we saved the file (in which segmentation was turned on) with another name as a working file (keeping the first file as a baseline to return to if needed). Don't ask me how, but without us knowing it, Sisulizer turned the setting off in the working file when it was saved with the new name, closed, and then reopened. If you check the setting before closing the file, it is still turned on.

I am uploading the file in question so that you can test saving it with another name yourself. Be sure to close the file after saving it with the new name using the menu option "File=>Save as", and then reopen it. The setting should be changed. Both myself and my colleague have confirmed it by recreating the situation several times on different computers. We even tested making some changes in the file before closing it. The setting was still changed when we opened it again.

When the project was then scanned for changes which should only have been a few, to our frustration, 1500+ changes were again indicated unexpectedly, because the segmentation setting had been altered unknowingly and the paragraphs that Sisulizer had previously seen as single segments were now broken up into the separate sentences. I have a feeling that this may also be the reason for strings that hadn't actually changed to be marked that way.

We only found this because we inadvertently checked the segmentation setting again in the working file discovering that it had been changed. It was then we were able to go back to the file I'm sending you and recreate the situation.

Regards,

Allen

Back To Top PM Private Upload Quote Reply

 Posted: Wed Jul 7th, 2010 06:49 pm
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
Hello Allen,

I am uploading the file in question so that you can test saving it with another name yourself. Be sure to close the file after saving it with the new name using the menu option "File=>Save as", and then reopen it. The setting should be changed. Both myself and my colleague have confirmed it by recreating the situation several times on different computers. We even tested making some changes in the file before closing it. The setting was still changed when we opened it again.
Thank you for this information. I also reproduced this behavior with your project.

I think we may have finally found the root of all evil in this issue which has caused us literally weeks of work.
Personally, I think you can be frustrated and I not to blame you, because here overlapped are 3 different issue:
- Sisulizer turn off/on segmentation rules for items with changed internal sub-tags structure (I described this issue above).
- Sisulizer turn off segmentation's using for project, if you use "Save as" feature.
- Adobe often change rules of code creation in RoboHelp. So, even if you don't change contents of HTML files, RoboHelp can do it (after upgrade to newer version/build) without your knowledge. Of course, Sisulizer detect these updated items as changed. As I mentioned in my first post of this thread, our developers hard work on improving of context support for HTML platform, but this is really difficult for platforms without real context (e.g. HTML).

Best regards,
Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

 Posted: Fri Jul 16th, 2010 10:36 am
PM Private Upload Quote Reply
Janusz Grzybek
Super Moderator


Joined: Fri Dec 1st, 2006
Location: Zabrze, Poland
Posts: 2660
Status: 
Offline
Hi Allen,

- Sisulizer turn off segmentation's using for project, if you use "Save as" feature.
Fixed in build 307. New build should come out in next week.

Janusz



____________________
http://www.sisulizer.com - Three simple steps to localize
Back To Top PM Private Upload Quote Reply

Current time is 05:15 pm  
.NET, Delphi, ... - Sisulizer Localization Tool Support > Technical Support (You need to be registered at the forum to write) > Bugs and Quirks in Sisulizer > Help topics "changed" incorrectly after scanning of new help file



WowUltra modified by Sisulizer Copyright © 2007-09 by Jim Hale - Based on WowBB Copyright © 2003-2006 Aycan Gulez

Sisulizer software localization tool - Three simple steps to localize