Re: Complex Script, BIDI text supported?
Just committed a fix for relative fo:block-container indents in bidi context at http://github.com/skynavga/fop. I think this takes care of all the issue you have reported. Let me know if you run into further problems. G. On Sun, Nov 27, 2011 at 7:51 AM, Glenn Adams gl...@skynav.com wrote: Hi Matthias, I've updated my working repo git://github.com/skynavga/fop.git with a fix for fo:leader to account for RTL writing modes. Let me know if you have any problems. I'm now working on a fix for fo:block-container that uses non-absolute positioning in a bidi context. Regards, Glenn On Wed, Mar 30, 2011 at 9:43 PM, Glenn Adams gl...@skynav.com wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre
Re: Complex Script, BIDI text supported?
Hi Matthias, I've updated my working repo git://github.com/skynavga/fop.git with a fix for fo:leader to account for RTL writing modes. Let me know if you have any problems. I'm now working on a fix for fo:block-container that uses non-absolute positioning in a bidi context. Regards, Glenn On Wed, Mar 30, 2011 at 9:43 PM, Glenn Adams gl...@skynav.com wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport area generated by block-container. In fact there is no relationship between WM and reference orientation (and thus CTM on
Re: Complex Script, BIDI text supported?
See also http://github.com/skynavga/fop. On Sun, Nov 27, 2011 at 7:51 AM, Glenn Adams gl...@skynav.com wrote: Hi Matthias, I've updated my working repo git://github.com/skynavga/fop.git with a fix for fo:leader to account for RTL writing modes. Let me know if you have any problems. I'm now working on a fix for fo:block-container that uses non-absolute positioning in a bidi context. Regards, Glenn On Wed, Mar 30, 2011 at 9:43 PM, Glenn Adams gl...@skynav.com wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport
Re: Complex Script, BIDI text supported?
Matthias et al., The following issues have been fixed, and are available in the i18n.arabic branch at http://github.com/skynavga/fop: - fo:character in RTL context ( http://skynav.trac.cvsdude.com/fop/ticket/39) - fo:page-number-citation in RTL context ( http://skynav.trac.cvsdude.com/fop/ticket/40) Note that there remains an issue with fo:character in the context of certain scripts, e.g., Arabic, Indic, etc., in which case an fo:character boundary prevents script shaping (contextual variants, ligatures, etc) from occurring across the element boundary. I am investigating ways to resolve this, but it looks like it may require significant work to resolve. I am tracking this separate problem at http://skynav.trac.cvsdude.com/fop/ticket/35. Regards, Glenn 2011/5/16 Matthias Reischenbacher matthias8...@gmx.at Hi Glenn, attached you'll find a test case for the three reported problems. I hate to ask, but could you please fix issue # 1 (fo:page-number-citation) soon? That would be of great help. Thanks! Matthias On 13.05.2011 11:36, Glenn Adams wrote: Matthias, Thanks. Yes, it would be helpful if you could send me minimal test cases for these issues, so I can address them. Regards, Glenn On Fri, May 13, 2011 at 5:58 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias
RE: Complex Script, BIDI text supported?
Once I have the normal one set up I will be looking at the bidi (we need it for Arabic) the one sticking point for us is we need the page numbering reversed on the fo:page-number/ and also for a multi column document we need it to go Right to left, does the git version have this? Or is that still in development? Theresa Forster From: Matthias Reischenbacher [mailto:matthias8...@gmx.at] Sent: 13 May 2011 12:58 To: fop-users@xmlgraphics.apache.org Subject: Re: Complex Script, BIDI text supported? Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: * fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; * fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G
Re: Complex Script, BIDI text supported?
Matthias, Thanks. Yes, it would be helpful if you could send me minimal test cases for these issues, so I can address them. Regards, Glenn On Fri, May 13, 2011 at 5:58 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre
Re: Complex Script, BIDI text supported?
Theresa, The current code in git://github.com/skynavga/fop.git DOES handle right-to-left column progression in fo:table. You need to ensure that the appropriate writing mode is specified on fo:table or on the nearest ancestor fo:block-container or on fo:page-sequence. I have NOT yet implemented changes needed to reverse page numbering, but it is on my list of TODO items. Thanks for reminding me. Regards, Glenn On Fri, May 13, 2011 at 7:26 AM, Theresa Jayne Forster ther...@inbrand.co.uk wrote: Once I have the normal one set up I will be looking at the bidi (we need it for Arabic) the one sticking point for us is we need the page numbering reversed on the fo:page-number/ and also for a multi column document we need it to go Right to left, does the git version have this? Or is that still in development? Theresa Forster *From:* Matthias Reischenbacher [mailto:matthias8...@gmx.at] *Sent:* 13 May 2011 12:58 *To:* fop-users@xmlgraphics.apache.org *Subject:* Re: Complex Script, BIDI text supported? Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line
RE: Complex Script, BIDI text supported?
Sorry if I didn’t put it in terms you could easily understand but I am making a variable 30+ page document with Flow and columns=”2” so not using tables at all. The generic flow of a page needs to be from “as we see it” back to front – right to left so on a 2 column page it needs to go right hand column to left Theresa From: Glenn Adams [mailto:gl...@skynav.com] Sent: 13 May 2011 15:40 To: fop-users@xmlgraphics.apache.org Subject: Re: Complex Script, BIDI text supported? Theresa, The current code in git://github.com/skynavga/fop.git DOES handle right-to-left column progression in fo:table. You need to ensure that the appropriate writing mode is specified on fo:table or on the nearest ancestor fo:block-container or on fo:page-sequence. I have NOT yet implemented changes needed to reverse page numbering, but it is on my list of TODO items. Thanks for reminding me. Regards, Glenn On Fri, May 13, 2011 at 7:26 AM, Theresa Jayne Forster ther...@inbrand.co.uk wrote: Once I have the normal one set up I will be looking at the bidi (we need it for Arabic) the one sticking point for us is we need the page numbering reversed on the fo:page-number/ and also for a multi column document we need it to go Right to left, does the git version have this? Or is that still in development? Theresa Forster From: Matthias Reischenbacher [mailto:matthias8...@gmx.at] Sent: 13 May 2011 12:58 To: fop-users@xmlgraphics.apache.org Subject: Re: Complex Script, BIDI text supported? Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts.. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: * fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; * fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G.. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align
Re: Complex Script, BIDI text supported?
Ah, sorry, was reading too quickly. Column progression in a page does indeed NOT yet make use of writing mode. I will add this to my defect/feature tracking list for this work, which I am in the process of updating now. Regards, Glenn On Fri, May 13, 2011 at 8:49 AM, Theresa Jayne Forster ther...@inbrand.co.uk wrote: Sorry if I didn’t put it in terms you could easily understand but I am making a variable 30+ page document with Flow and columns=”2” so not using tables at all. The generic flow of a page needs to be from “as we see it” back to front – right to left so on a 2 column page it needs to go right hand column to left Theresa *From:* Glenn Adams [mailto:gl...@skynav.com] *Sent:* 13 May 2011 15:40 *To:* fop-users@xmlgraphics.apache.org *Subject:* Re: Complex Script, BIDI text supported? Theresa, The current code in git://github.com/skynavga/fop.git DOES handle right-to-left column progression in fo:table. You need to ensure that the appropriate writing mode is specified on fo:table or on the nearest ancestor fo:block-container or on fo:page-sequence. I have NOT yet implemented changes needed to reverse page numbering, but it is on my list of TODO items. Thanks for reminding me. Regards, Glenn On Fri, May 13, 2011 at 7:26 AM, Theresa Jayne Forster ther...@inbrand.co.uk wrote: Once I have the normal one set up I will be looking at the bidi (we need it for Arabic) the one sticking point for us is we need the page numbering reversed on the fo:page-number/ and also for a multi column document we need it to go Right to left, does the git version have this? Or is that still in development? Theresa Forster *From:* Matthias Reischenbacher [mailto:matthias8...@gmx.at] *Sent:* 13 May 2011 12:58 *To:* fop-users@xmlgraphics.apache.org *Subject:* Re: Complex Script, BIDI text supported? Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts.. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G.. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align
Re: Complex Script, BIDI text supported?
You can track the status of these defects at: - right-to-left page numbering - http://skynav.trac.cvsdude.com/fop/ticket/33 - right-to-left flow column progression - http://skynav.trac.cvsdude.com/fop/ticket/34 Regards, Glenn On Fri, May 13, 2011 at 9:27 AM, Glenn Adams gl...@skynav.com wrote: Ah, sorry, was reading too quickly. Column progression in a page does indeed NOT yet make use of writing mode. I will add this to my defect/feature tracking list for this work, which I am in the process of updating now. Regards, Glenn On Fri, May 13, 2011 at 8:49 AM, Theresa Jayne Forster ther...@inbrand.co.uk wrote: Sorry if I didn’t put it in terms you could easily understand but I am making a variable 30+ page document with Flow and columns=”2” so not using tables at all. The generic flow of a page needs to be from “as we see it” back to front – right to left so on a 2 column page it needs to go right hand column to left Theresa *From:* Glenn Adams [mailto:gl...@skynav.com] *Sent:* 13 May 2011 15:40 *To:* fop-users@xmlgraphics.apache.org *Subject:* Re: Complex Script, BIDI text supported? Theresa, The current code in git://github.com/skynavga/fop.git DOES handle right-to-left column progression in fo:table. You need to ensure that the appropriate writing mode is specified on fo:table or on the nearest ancestor fo:block-container or on fo:page-sequence. I have NOT yet implemented changes needed to reverse page numbering, but it is on my list of TODO items. Thanks for reminding me. Regards, Glenn On Fri, May 13, 2011 at 7:26 AM, Theresa Jayne Forster ther...@inbrand.co.uk wrote: Once I have the normal one set up I will be looking at the bidi (we need it for Arabic) the one sticking point for us is we need the page numbering reversed on the fo:page-number/ and also for a multi column document we need it to go Right to left, does the git version have this? Or is that still in development? Theresa Forster *From:* Matthias Reischenbacher [mailto:matthias8...@gmx.at] *Sent:* 13 May 2011 12:58 *To:* fop-users@xmlgraphics.apache.org *Subject:* Re: Complex Script, BIDI text supported? Hi Glenn, I've tested your git branch and the following additional issues came up: 1. fo:basic-links with internal destination don't work 2. fo:page-number references only sometimes work 3. start-indent on fo:block with a block-container child isn't taken into account. Please let me know, if you want me to prepare a test case showing all three issues. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts.. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G.. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region
Re: Complex Script, BIDI text supported?
Hi Glenn, thanks for your status update! I will test your changes in the next fiew days. I'll need support for fo:leaders, so thanks for taking care of that too. Best regards, Matthias On 31.03.2011 00:43, Glenn Adams wrote: Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: * fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; * fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com mailto:gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at mailto:matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of
Re: Complex Script, BIDI text supported?
Matthias, I just updated my working repo git://github.com/skynavga/fop.git with fixes for fo:table and fo:list-block to account for RTL writing modes; i.e., table column progression and list-item (label and body) alignment are now RTL in RTL WM contexts. In the process, I uncovered two other issues as well, which I am now working on along with the fo:character problem you reported: - fo:block-container does not align right in a RTL WM context when it does not use absolute or fixed positioning; - fo:leader is not working in a RTL WM context; I will update you when I have fixes committed for these issues. G. On Wed, Mar 23, 2011 at 6:45 AM, Glenn Adams gl...@skynav.com wrote: Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport area generated by block-container. In fact there is no relationship between WM and reference orientation (and thus CTM on viewport area). I'll let you know when I've made the commit. G. On Wed, Mar 9, 2011 at 2:10 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Ok, thanks a lot for your help! Matthias Glenn Adams-2 wrote: It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the
Re: Complex Script, BIDI text supported?
Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period with fo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adams gl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport area generated by block-container. In fact there is no relationship between WM and reference orientation (and thus CTM on viewport area). I'll let you know when I've made the commit. G. On Wed, Mar 9, 2011 at 2:10 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Ok, thanks a lot for your help! Matthias Glenn Adams-2 wrote: It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the writing-mode and (default) font-family on fo:page-sequence, it should produce correct results. See attached for what I'm getting (using the current GITHUB repo) with a modified FO. I would advise you use the GIT repository for now, because it is kept up to date with my working copy. I submit a patch irregularly for integration into the SVN Temp_ComplexScripts branch. G. On Wed, Mar 9, 2011 at 12:21 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on git hub? I've never used git therefore I first went with the apache SVN branch. Thanks Regards, Matthias Glenn Adams-2 wrote: Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at git://github.com/skynavga/fop.git. Regards, Glenn On Wed, Mar 9, 2011 at 11:28 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode
Re: Complex Script, BIDI text supported?
Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport area generated by block-container. In fact there is no relationship between WM and reference orientation (and thus CTM on viewport area). I'll let you know when I've made the commit. G. On Wed, Mar 9, 2011 at 2:10 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Ok, thanks a lot for your help! Matthias Glenn Adams-2 wrote: It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the writing-mode and (default) font-family on fo:page-sequence, it should produce correct results. See attached for what I'm getting (using the current GITHUB repo) with a modified FO. I would advise you use the GIT repository for now, because it is kept up to date with my working copy. I submit a patch irregularly for integration into the SVN Temp_ComplexScripts branch. G. On Wed, Mar 9, 2011 at 12:21 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on git hub? I've never used git therefore I first went with the apache SVN branch. Thanks Regards, Matthias Glenn Adams-2 wrote: Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at
Re: Complex Script, BIDI text supported?
Thanks for uncovering these issues. Let me address them and I'll let you know when they are fixed in my GIT repo. Perhaps after that I can submit a new patch update for the SVN branch. G. 2011/3/23 Matthias Reischenbacher matthias8...@gmx.at Glenn, I just tested a little bit more and the following questions/problems came up: 1. Are table cells also supposed to flow from right to left in rl-tb writing-mode? 2. Shouldn't list-item labels be displayed on the right side? 3. fo:character seem to overlap in rl mode. I'm attaching a new test file and the current PDF output. Thanks, Matthias On 22.03.2011 18:27, Matthias Reischenbacher wrote: Hi Glenn, thanks again for your help and doing all those fixes. Today I've retested my sample file with the git branch and everything works fine now. I'll do some more testing in the next weeks and I'll let you know if I encounter any problems. Are you planning to submit a new patch for the apache svn repo anytime soon? If not, no problem. I'm just asking because I have some local modifications in my svn working copy which I'll have to apply manually to the git working copy. Thanks best regards, Matthias Glenn Adams-2 wrote: Matthias, I've fixed the fo:block-container problem as well as text-align and treatment of writing-mode on page regions. The fixes are present in my GIT repo I referred to earlier (and also include all trunk commits up to this point). To be a little more detailed: - the use of writing-mode=rl on simple-page-master now results in correct interpretation of region-start and region end; i.e., region-start is now on the right side of the page (with 0 reference orientation), and region-end is on the left side; - the erroneous (historical) mirroring around the y axis of viewports with writing-mode=rl, including those generated by region-* and block-container, has been removed; - both text-align and text-align-last now treat start and end as relative to the writing mode, e.g., text-align=start is on the right when writing-mode=rl, etc. - if no text-align is specified, then in a writing-mode=rl context correctly results in right (i.e., start) alignment, while in lr start is on left (assuming reference orientation of 0); I am aware of an existing problem with text-align=justify in RTL WM contexts, and expect a fix in a day or so. Let me know if you encounter any other bidi problems. Also, just as a reminder, a neutral character, e.g., FULL STOP '.', at the end of a paragraph in a block takes on the default direction of the block. So, if you have a block of English text with a period at the end and it appears in a RTL WM context, then the period will be placed at the left of the last line, not the right as one might expect. This is actually correct behavior, and is a side effect of the Unicode bidi algorithm. To deal with this, you can wrap the English text including the period withfo:bidi-override unicode-bidi=embed direction=ltrSentence(s) with a period at the end./fo:bidi-override. This will create a LTR embedding context in the outer RTL context of the block. In general, you can use the fo:bidi-override element to both create bidi embedding levels and also to override the directionality of text as would be computed by the Unicode bidi algorithm. G. On Wed, Mar 9, 2011 at 2:28 PM, Glenn Adamsgl...@skynav.com wrote: I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport area generated by block-container. In fact there is no relationship between WM and reference orientation (and thus CTM on viewport area). I'll let you know when I've made the commit. G. On Wed, Mar 9, 2011 at 2:10 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Ok, thanks a lot for your help! Matthias Glenn Adams-2 wrote: It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the writing-mode and (default) font-family on fo:page-sequence, it should produce correct results. See attached for what I'm getting (using the current GITHUB repo) with a modified FO. I would advise you use the GIT repository for now, because it is kept up to date with my working copy. I submit a patch irregularly for integration into the SVN Temp_ComplexScripts branch. G. On Wed, Mar 9, 2011 at 12:21 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on
Complex Script, BIDI text supported?
Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode MS font and writing-mode set to rl-tb. The Hebrew text is displayed fine but english characters are also mirrored. Do I have to set an additional xsl-fo attribute in order to correctly display english characters or is the BIDI text algorithm not yet implemented? Thanks for your help, Matthias Reischenbacher -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109094.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
Re: Complex Script, BIDI text supported?
Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at git://github.com/skynavga/fop.git. Regards, Glenn On Wed, Mar 9, 2011 at 11:28 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode MS font and writing-mode set to rl-tb. The Hebrew text is displayed fine but english characters are also mirrored. Do I have to set an additional xsl-fo attribute in order to correctly display english characters or is the BIDI text algorithm not yet implemented? Thanks for your help, Matthias Reischenbacher -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109094.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
Re: Complex Script, BIDI text supported?
Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on git hub? I've never used git therefore I first went with the apache SVN branch. Thanks Regards, Matthias Glenn Adams-2 wrote: Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at git://github.com/skynavga/fop.git. Regards, Glenn On Wed, Mar 9, 2011 at 11:28 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode MS font and writing-mode set to rl-tb. The Hebrew text is displayed fine but english characters are also mirrored. Do I have to set an additional xsl-fo attribute in order to correctly display english characters or is the BIDI text algorithm not yet implemented? Thanks for your help, Matthias Reischenbacher -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109094.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109566.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
Re: Complex Script, BIDI text supported?
It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the writing-mode and (default) font-family on fo:page-sequence, it should produce correct results. See attached for what I'm getting (using the current GITHUB repo) with a modified FO. I would advise you use the GIT repository for now, because it is kept up to date with my working copy. I submit a patch irregularly for integration into the SVN Temp_ComplexScripts branch. G. On Wed, Mar 9, 2011 at 12:21 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on git hub? I've never used git therefore I first went with the apache SVN branch. Thanks Regards, Matthias Glenn Adams-2 wrote: Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at git://github.com/skynavga/fop.git. Regards, Glenn On Wed, Mar 9, 2011 at 11:28 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode MS font and writing-mode set to rl-tb. The Hebrew text is displayed fine but english characters are also mirrored. Do I have to set an additional xsl-fo attribute in order to correctly display english characters or is the BIDI text algorithm not yet implemented? Thanks for your help, Matthias Reischenbacher -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109094.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109566.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org ?xml version=1.0 encoding=UTF-8? fo:root xmlns:fo=http://www.w3.org/1999/XSL/Format; fo:layout-master-set fo:simple-page-master master-name=simple page-height=5in page-width=5in fo:region-body/ /fo:simple-page-master /fo:layout-master-set fo:page-sequence master-reference=simple writing-mode=rl-tb font-family=Arial Unicode MS fo:flow flow-name=xsl-region-body fo:blockחוברת הוראות בטיחות זו מתמקדת בכמה מהמצבים העקרוניים שעשויים להתרחש בעת הפעלת המכונה ותחזוקתה./fo:block fo:blockTest/fo:block /fo:flow /fo:page-sequence /fo:root test.pdf Description: Adobe PDF document - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
Re: Complex Script, BIDI text supported?
Ok, thanks a lot for your help! Matthias Glenn Adams-2 wrote: It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the writing-mode and (default) font-family on fo:page-sequence, it should produce correct results. See attached for what I'm getting (using the current GITHUB repo) with a modified FO. I would advise you use the GIT repository for now, because it is kept up to date with my working copy. I submit a patch irregularly for integration into the SVN Temp_ComplexScripts branch. G. On Wed, Mar 9, 2011 at 12:21 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on git hub? I've never used git therefore I first went with the apache SVN branch. Thanks Regards, Matthias Glenn Adams-2 wrote: Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at git://github.com/skynavga/fop.git. Regards, Glenn On Wed, Mar 9, 2011 at 11:28 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode MS font and writing-mode set to rl-tb. The Hebrew text is displayed fine but english characters are also mirrored. Do I have to set an additional xsl-fo attribute in order to correctly display english characters or is the BIDI text algorithm not yet implemented? Thanks for your help, Matthias Reischenbacher -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109094.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109566.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org ?xml version=1.0 encoding=UTF-8? fo:root xmlns:fo=http://www.w3.org/1999/XSL/Format; fo:layout-master-set fo:simple-page-master master-name=simple page-height=5in page-width=5in fo:region-body/ /fo:simple-page-master /fo:layout-master-set fo:page-sequence master-reference=simple writing-mode=rl-tb font-family=Arial Unicode MS fo:flow flow-name=xsl-region-body fo:blockחוברת הוראות בטיחות זו מתמקדת בכמה מהמצבים העקרוניים שעשויים להתרחש בעת הפעלת המכונה ותחזוקתה./fo:block fo:blockTest/fo:block /fo:flow /fo:page-sequence /fo:root - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31110436.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
Re: Complex Script, BIDI text supported?
I have a fix for the block-container issue as well, which I will commit shortly to my GIT repo. The problem was that earlier (pre complex script) code apparently assumed that right-to-left writing mode was handled by mirroring the CTM around the y axis, so it was setting the CTM incorrectly for the viewport area generated by block-container. In fact there is no relationship between WM and reference orientation (and thus CTM on viewport area). I'll let you know when I've made the commit. G. On Wed, Mar 9, 2011 at 2:10 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Ok, thanks a lot for your help! Matthias Glenn Adams-2 wrote: It appears the problem is due to a bug related to the use of block-container, which I will look into. However, if you remove block-container, and put the writing-mode and (default) font-family on fo:page-sequence, it should produce correct results. See attached for what I'm getting (using the current GITHUB repo) with a modified FO. I would advise you use the GIT repository for now, because it is kept up to date with my working copy. I submit a patch irregularly for integration into the SVN Temp_ComplexScripts branch. G. On Wed, Mar 9, 2011 at 12:21 PM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, here it goes: Fo file: http://old.nabble.com/file/p31109566/hebrew_test.xml hebrew_test.xml PDF file: http://old.nabble.com/file/p31109566/hebrew_test.pdf hebrew_test.pdf Should I try building your code on git hub? I've never used git therefore I first went with the apache SVN branch. Thanks Regards, Matthias Glenn Adams-2 wrote: Matthias, Thanks for the report. Could you send me the FO input file and PDF output file? I am actively testing and fixing the Bidi and Script support, with real-time updates occurring at git://github.com/skynavga/fop.git. Regards, Glenn On Wed, Mar 9, 2011 at 11:28 AM, Matthias Reischenbacher matthias8...@gmx.at wrote: Glenn, I tried your work today of the complex script branch. I managed to generate a PDF file with Hebrew text using Arial Unicode MS font and writing-mode set to rl-tb. The Hebrew text is displayed fine but english characters are also mirrored. Do I have to set an additional xsl-fo attribute in order to correctly display english characters or is the BIDI text algorithm not yet implemented? Thanks for your help, Matthias Reischenbacher -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109094.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31109566.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org ?xml version=1.0 encoding=UTF-8? fo:root xmlns:fo=http://www.w3.org/1999/XSL/Format; fo:layout-master-set fo:simple-page-master master-name=simple page-height=5in page-width=5in fo:region-body/ /fo:simple-page-master /fo:layout-master-set fo:page-sequence master-reference=simple writing-mode=rl-tb font-family=Arial Unicode MS fo:flow flow-name=xsl-region-body fo:blockחוברת הוראות בטיחות זו מתמקדת בכמה מהמצבים העקרוניים שעשויים להתרחש בעת הפעלת המכונה ותחזוקתה./fo:block fo:blockTest/fo:block /fo:flow /fo:page-sequence /fo:root - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org -- View this message in context: http://old.nabble.com/Complex-Script%2C-BIDI-text-supported--tp31109094p31110436.html Sent from the FOP - Users mailing list archive at Nabble.com. - To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org