Apache OpenOffice Wiki - User contributions [en]

Education Project/Effort/Math baseline alignment/Debuging starmath

2009-01-24T07:15:29Z

RakeshPandit:

=== <div align="left" style="color:blue; font-size:18px">'''Milestone '''</div> ===

Current Build is m245, but m247 should work too

install it wherever ( Desktop is a good option on Mac OS X)

===Debug ===

To trace, analyze, understand the issue and fix it, it is mandatory to debug starmath.

For that purpose, we will use the most known tool : [http://sourceware.org/gdb/ ''the GNU Project debugger, aka gdb'''] ( included in XCode on Mac OS X ).

For the one discovering '''gdb''', there is [http://sources.redhat.com/gdb/current/onlinedocs/gdb_toc.html '''online gdb tutorial'''] or [http://www.cs.cmu.edu/~gilpin/tutorial/ another one there ]

==== What is the problem ? ====
all libmath are stripped by default, and you'll have to add the symbols for debugging. Thus you'll have to rebuild starmath, and symlink with the new math libs.

The problem is, since sb71 integration, use symlinks with libs is just a workaround. See [[http://www.openoffice.org/issues/show_bug.cgi?id=83548 '''issue 83548''' for more information ]]

'''The clean solution : replace libstarmaths* in the bundle after every rebuild'''

Todo : write a little shell script doing the job ;-)

==== (workaround) A patch for linkoo ====

Waiting, the following patch (thanks to Kendy !! ) should allow linkoo to work as expected : [http://eric.bachard.free.fr/Education/february2008/linkoo/fix-linkoo_new.diff patch for linkoo]

==== (workaround) Apply the patch ====

Just put it in the root dir ( the one containing all OpenOffice.org sources), and do :

1) fake the patch application ( using --dry-run option )

patch --dry-run -p0 <fix-linkoo_new.diff

2) if nothing bad occurs (excepted some fuzz) apply it for true :

patch -p0 <fix-linkoo_new.diff

With this patch applied, linkoo should work as expected. Please note this is a workaround, to make linkoo work for debug purpose.

===<div align="left" style="color:blue; font-size:18px">'''Process '''</div> ===

==== Add debug infos in starmath ====

1) Build adding rectangles (for baselines and co )

-> uncomment line 93 in node.cxx : all formulas will be drawn with colored lines giving all alignment information ( Top Bottom Baseline when existing ..etc ) -> See the screenshots below

(This line seems to me moved now. Corresponding SM_RECT_DEBUG is at line 59. But uncommenting wouldn't work there. Rather move this to inc/rect.hxx) -- rakesh *need to provide patch upstream later* It is [http://rakesh.fedorapeople.org/misc/starmath.patch here]

<pre>
Index: ./inc/rect.hxx
===================================================================
RCS file: /cvs/sw/starmath/inc/rect.hxx,v
retrieving revision 1.7
diff -u -r1.7 rect.hxx
--- ./inc/rect.hxx 24 Apr 2008 13:57:29 -0000 1.7
+++ ./inc/rect.hxx 15 Sep 2008 17:33:58 -0000
@@ -62,6 +62,7 @@
// Note: however, italic spaces can be negative!
//

+#define SM_RECT_DEBUG
// possible flags for the 'Draw' function below (just for debugging)
#define SM_RECT_CORE 0x0001
#define SM_RECT_ITALIC 0x0002
</pre>

==== Rebuild starmath libs ====

2) Build starmath again using :

cd starmath

mv unxmacxi.pro unxmacxi.pro_backup /* adapt to your architecture / OS */

build debug="something_but_not_empty"



==== Linkoo ====

If never done before, run linkoo. '''Important: one time only !!'''

=====Method =====

* Notations

BUILD_SRC_DIR is the directory containing all the sources

INSTALL_DIR is the location of the Bundle

command line

* Go into <BUILD_SRC_DIR>/solenv/bin

cd <BUILD_SRC_DIR>/solenv/bin

* Run linkoo ( fake command, to verify everything is ok )

--dry-run is for the simulation. don't forget it ( else you can screew your tree )

Please notice, that in our case :

INSTALL_DIR=~/Desktop/test_m247_aquavcl05

BUILD_SRC_DIR =~/Desktop/SRC680_m247

Thus, the complete command line is :

./linkoo ~/Desktop/test_m247_aquavcl05/OpenOffice.org.app/Contents/ ~/Desktop/SRC680_m247 --dry-run

<div align="left" style="color:red; font-size:18px">'''PLEASE READ ''' [[Education_Project/Effort/Math_baseline_alignment/Debuging_starmath/linkoo log | Linkoo Instalation log</div> ]]

* '''Run linkoo for true''' ( using --product option )

./linkoo ~/Desktop/test_m247_aquavcl05/OpenOffice.org.app/Contents/ ~/Desktop/SRC680_m247 --product

* Run ooenv

'''To avoid strange effects, once linkoo finished, please don't forget to source ooenv before ./soffice. '''

ooenv is located in '''INSTALL_DIR/OpenOffice.org.app/Contents/program''' (means ~/Desktop/test_m247_aquavcl05/OpenOffice.org.app/Contents/program )

./ooenv

And then launch gdb as described below

===== Mac OS X Linkoo Issue =====

'''FIXME''' : linkoo seems to not work ( I'll have a look asap )

Workaround :

'''Link the new libs ( or use linkoo ) from inside the bundle'''.

*Go into the Bundle :

cd ~/Desktop/test_m247_aquavcl05/OpenOffice.org.app/Contents/MacOS

*Remove the striped libs :

rm -f libsm680mxi.dylib libsmd680mxi.dylib

*create symlinks for the non strippied one :

ln -s ~/Desktop/SRC680_m247/starmath/unxmacxi.pro/lib/libsm680mxi.dylib .

ln -s ~/Desktop/SRC680_m247/starmath/unxmacxi.pro/lib/libsmd680mxi.dylib .

!! don't forget the points !!

==== Attach the running process to gdb ====

* Launch OpenOffice.org

Should work out of the box :)

* Find the processus number of soffice.bin

To find the correct processus number correponding to OpenOffice.org binary execution, just run the following command line in a terminal :

ps ax | grep soffice.bin | grep -v grep

For example:

ordinateur-de-eric-b-2:~ ericb$ ps ax | grep soffice.bin | grep -v grep

11666 ?? S 0:00.56 /Applications/OpenOffice.org 2.4.app/Contents/MacOS/soffice.bin -psn_0_51511297

-> the processus number is 11666



* In a terminal, run gdb

ordinateur-de-eric-b-2:~ ericb$ gdb -q

(gdb)

... and attach the process number corresponding to soffice.bin

<pre>
(gdb) attach 11666
Attaching to process 11666.
Reading symbols for shared libraries . done
Reading symbols for shared libraries .............................................................
................................................................................ done
0x90009bf7 in mach_msg_trap ()
(gdb)
</pre>

==== Complement : detach the process ====

Simply use detach :

<pre>
(gdb) detach
Detaching from process 11666 thread 0xd03.
(gdb)
</pre>

Other possibility :
<pre>
(gdb) q
The program is running. Quit anyway (and detach it)? (y or n) y
Detaching from process 11666 thread 0xd03.
ordinateur-de-eric-b-2:~ ericb$
</pre>

===Interesting breakpoints===

FIXME : non exhaustive list, to be completed with other examples asap

==== tracing all Arrange() methods in node.cxx ====

SmRectangleNode::Arrange ( starmath/source/node.cxx, line 2329 )

SmSubSupNode::Arrange ( starmath/source/node.cxx, line 1437 )

SmMathSymbolNode::Arrange ( starmath/source/node.cxx, line 2731 )

SmBinVerNode::Arrange ( starmath/source/node.cxx, line 1087 )

There are exactly 27 Arrange() methods in node.cxx, and all must be analyzed.

For the complete list, see : [[Education_Project/Effort/Math_baseline_alignment/Debuging_starmath/Arrange() methods list | Arrange methods list]]

==== Other interesting breakpoints ====

FIXME

=== gdb completion ===

As preliminary, don't forget to use gdb completion feature: it's great and very usefull when you have to choose between several possibilities, like the same but overloaded method ... etc.

Principle: write the first letters of the word you want to write, and when sure, hit TAB key ( hit it several times if needed )

* either there is only one solution, and the word will be simply completed (nice, isn't it ? )
* or several solution will be proposed, and you'll have to decide which one use ..

<br>

=== How to add breakpoints ? ===

<br>
A breakpoint is an extremely helpfull way to stop -repeatably and safely- the binary execution, and analyse what happens with the code, modify on the fly, test other values ..etc

Several possibilities can be used there, to add a breakpoint :

1) specify the line :

Insert a breakpoint line 1234 in node.cxx :

(gdb) b node.cxx:1234

2) specify the method:

Insert a breakpoint when a method is called :

(gdb) b 'SmBinVerNode::Arrange'

Note: when the same method is overloaded, you can add the arguments to be sure to hit the right method.

3) conditional break :

Insert a breakpoint when the condition is verified :

(gdb) b cond ( (thing == true) || (foo != 0) )

=== Examples ===

First try:

type an equation containing a^2 + b^2 , exit formula ( ESC twice ) then enter the word "text"

then create a new equation containing %tau a^2 + b^2 and notice the difference.

First breakpoint: parse.cxx:398

When entering whatever letter, you hit a breakpoint. Just trace after it, and compare the two equations (e.g.)

Entering %tau a^2 + b^2 , you'll see the line starmath/source/parse.cxx:770

*FIXME* : verify what makes ( if (aTmpRes.TokenType & KParseType::IDENTNAME)) true

===Screeenshots ===

[[Image:Issue972_tested_case05.jpg]]

[[Image:issue972_tested_case21.jpg]]

=== Links ===

A lot of screenshots and more informations are available at : [[http://eric.bachard.free.fr/Education/february2008/starmath/ Ericb's site starmath ]]

Education Project/Effort/Math baseline alignment/Mails

2008-12-06T08:48:40Z

RakeshPandit:

-----------------------------------

Hello Thomas,

I had been debugging starmath module. Using some test cases and
checking the flow.

There are some very basic doubts:

1. In my view centre(0,0) of co-ordinate system from which all
calculations are made (e.g calculation of Point, Alignment etc) is at
top left corner. The +ve side of Y axis goes down the left corner
towards bottom and +ve side of X axis goes horizontal towards top
right corned. May you confirm this ?

2. Now as per discussion which has happened so far, every Node needs
to have baseline calculated. To start with, I think I am ready to
write code for simple cases like "over". My strategy for over is
recursive calculation of baseline. After moving numerator to top and
denominator to bottom we will find the average of both and assign this
to baseline of whole fraction and all this will go on recursively
automatically. Does it sound good ?

something like(just a pseudo code ) :

if (pNum->HasBaseline() && pDenom->HasBaseline())
{
SetHasBaseline(TRUE);
SetBaseline((pNum->GetBaseline() + pDenom->GetBaseline()) / 2L);
}

That needs fraction to be a node. And is it really a node ? I think yes.

These calculations are in SmBinVerNode::Arrange method.

3. There are many messages and comments in French, and I would like to
start converting important ones as and when they come into English.
Would that be fine ? And can I do it in same CWS ?

4. Should I reassign issue to myself and create a CWS? It hasn't been
created yet.

5. Can baseline be negative ? I don't think so, but there are test
cases where baseline has value -<number>

Thanks

--
Cheers,
rakesh

-----------------------------------

Hello Rakesh,

> Hello Thomas, <br>
><br>
> I had been debugging starmath module. Using some test cases and<br>
> checking the flow.<br>
><br>
> There are some very basic doubts:<br>
><br>
> 1. In my view centre(0,0) of co-ordinate system from which all<br>
> calculations are made (e.g calculation of Point, Alignment etc) is at<br>
> top left corner. The +ve side of Y axis goes down the left corner<br>
> towards bottom and +ve side of X axis goes horizontal towards top<br>
> right corned. May you confirm this ?<br>
><br>

Yes. That is not what one might expect, but for all window based
coordinates in most UI API's it is the same.

> 2. Now as per discussion which has happened so far, every Node needs<br>
> to have baseline calculated. To start with, I think I am ready to<br>
> write code for simple cases like "over". My strategy for over is<br>
> recursive calculation of baseline. After moving numerator to top and<br>
> denominator to bottom we will find the average of both and assign this<br>
> to baseline of whole fraction and all this will go on recursively<br>
> automatically. Does it sound good ?<br>
><br>
> something like(just a pseudo code ) :
>
> if (pNum->HasBaseline() && pDenom->HasBaseline())
> {
> SetHasBaseline(TRUE);
> SetBaseline((pNum->GetBaseline() +
> pDenom->GetBaseline()) / 2L);
> }
>
> That needs fraction to be a node. And is it really a node ? I think yes.<br>
><br>
> These calculations are in SmBinVerNode::Arrange method.<br>
><br>

The basic idea is not quite right. If formulas are layouted it is not<br>
about finding the overall height and setting the mid as baseline.<br>
Layout of fractions should be done in away that the horizontal line is<br>
exactly on the same y position as a "-" in "a-b" would be if there were<br>
further formula parts.<br>

See for example: a - {a over b} over {y+z} or a - {H^{2^3}} over {H+H}<br>
And compare with any book about physics or math.<br>

> 3. There are many messages and comments in French, and I would like to<br>
> start converting important ones as and when they come into English.<br>
> Would that be fine ? And can I do it in same CWS ?<br>
><br>
<br>
Err... Where are messages and comments in French?<br>
And do you want to put them into a CWS once you have them translated<br>
into English?<br>

> 4. Should I reassign issue to myself and create a CWS? It hasn't been<br>
> created yet.<br>
><br>
<br>

If you feel like it you can do so.
That would be a good idea if you want to share the code with otehrs e.g.
Eric or me while developing.
If you want that just drop me a note and I will setup a CWS. That is
because in the end our QA will have a look at it and it is easier to get
a build done/installed from here if the CWS is on our servers...

> 5. Can baseline be negative ? I don't think so, but there are test<br>
> cases where baseline has value -<number><br>
><br>

Yes it can become negative.
The actual baseline of the very outer line (or to put it different: of
the first char in a line) should not be negative. But it wouldn't matter.

The reason for baselines to become negative is because of aligning
different nodes vertically. If you take the following two formulas for
example
x - x
and
stack{a#b#c#d#e#f#g}
each on it's own none of the baselines should be negative.
But if they are on the same line like
x - x stack{a#b#c#d#e#f#g}
things may become different.
The reason for this is that the alignment of the whole formula starts
with the aligned node. And here that is 'x'. If, just for the example,
we assume a lines height will be 100 then the rectangle for 'a' might be
at pos 0,0 with size 60x100.
And the stack if placed at (0,0) would have a size of about 160x700.
If the latter one is to be vertically aligned to the first it is done by
moving it up until in order for the baseline of the stack{...} command
to match the one from "x-y". By doing that we surely will have the
baselines of a, b and c be moved into the negative y axis as can be seen
in the layouted formula.

But that doesn't really matter when the formula is displayed since it
will work with any logically position calculated.
At most one needs to move all the formula nodes in order to have the
whole formula start at pos (0,0) if needs be...

Thomas

-------------------------------------

2008/9/29 Thomas Lange - Sun Germany - ham02 - Hamburg <Thomas.Lange@sun.com>:<br>
[..]<br>
- Show quoted text -<br>
>> There are some very basic doubts:<br>
>><br>
>> 1. In my view centre(0,0) of co-ordinate system from which all<br>
>> calculations are made (e.g calculation of Point, Alignment etc) is at<br>
>> top left corner. The +ve side of Y axis goes down the left corner<br>
>> towards bottom and +ve side of X axis goes horizontal towards top<br>
>> right corned. May you confirm this ?<br>
>><br>
> Yes. That is not what one might expect, but for all window based<br>
> coordinates in most UI API's it is the same.<br>
><br>
>> 2. Now as per discussion which has happened so far, every Node needs<br>
>> to have baseline calculated. To start with, I think I am ready to<br>
>> write code for simple cases like "over". My strategy for over is<br>
>> recursive calculation of baseline. After moving numerator to top and<br>
>> denominator to bottom we will find the average of both and assign this<br>
>> to baseline of whole fraction and all this will go on recursively<br>
>> automatically. Does it sound good ?<br>
>><br>
>> something like(just a pseudo code ) :<br>
>><br>
>> if (pNum->HasBaseline() && pDenom->HasBaseline())<br>
>> {<br>
>> SetHasBaseline(TRUE);<br>
>> SetBaseline((pNum->GetBaseline() +<br>
>> pDenom->GetBaseline()) / 2L);<br>
>> }<br>
>><br>
>> That needs fraction to be a node. And is it really a node ? I think yes.<br>
>><br>
>> These calculations are in SmBinVerNode::Arrange method.<br>
>><br>
><br>
> The basic idea is not quite right. If formulas are layouted it is not<br>
> about finding the overall height and setting the mid as baseline.<br>
> Layout of fractions should be done in away that the horizontal line is<br>
> exactly on the same y position as a "-" in "a-b" would be if there were<br>
> further formula parts.<br>
><br>
> See for example: a - {a over b} over {y+z} or a - {H^{2^3}} over {H+H}<br>
> And compare with any book about physics or math.<br>
><br>
<br>
So, for simple cases like "<X> over <Y>" baseline would simply be
baseline of "over", if I am getting it right.

>> 3. There are many messages and comments in French, and I would like to<br>
>> start converting important ones as and when they come into English.<br>
>> Would that be fine ? And can I do it in same CWS ?<br>
>><br>
> Err... Where are messages and comments in French?<br>
> And do you want to put them into a CWS once you have them translated<br>
> into English?<br>
><br>

Actually they where German and few where in French (I am not sure
though, don't know either of them :( . I plan to start a page in wiki
and make a table with two columns. First one will contain messages in
non-English and other side will contain translated messges. For
critical comments related to this issue I will like to commit them in
this CWS.

>> 4. Should I reassign issue to myself and create a CWS? It hasn't been
>> created yet.<br>
>><br>
> If you feel like it you can do so.<br>
> That would be a good idea if you want to share the code with otehrs e.g.<br>
> Eric or me while developing.<br>
> If you want that just drop me a note and I will setup a CWS. That is<br>
> because in the end our QA will have a look at it and it is easier to get<br>
> a build done/installed from here if the CWS is on our servers...<br>
><br>

Okay.
I will do it, thanks
- Show quoted text -

>> 5. Can baseline be negative ? I don't think so, but there are test<br>
>> cases where baseline has value -<number><br>
>><br>
> Yes it can become negative.<br>
> The actual baseline of the very outer line (or to put it different: of<br>
> the first char in a line) should not be negative. But it wouldn't matter.<br>
><br>
> The reason for baselines to become negative is because of aligning<br>
> different nodes vertically. If you take the following two formulas for<br>
> example<br>
> x - x<br>
> and<br>
> stack{a#b#c#d#e#f#g}<br>
> each on it's own none of the baselines should be negative.<br>
> But if they are on the same line like<br>
> x - x stack{a#b#c#d#e#f#g}<br>
> things may become different.<br>
> The reason for this is that the alignment of the whole formula starts<br>
> with the aligned node. And here that is 'x'. If, just for the example,<br>
> we assume a lines height will be 100 then the rectangle for 'a' might be<br>
> at pos 0,0 with size 60x100.<br>
> And the stack if placed at (0,0) would have a size of about 160x700.<br>
> If the latter one is to be vertically aligned to the first it is done by<br>
> moving it up until in order for the baseline of the stack{...} command<br>
> to match the one from "x-y". By doing that we surely will have the<br>
> baselines of a, b and c be moved into the negative y axis as can be seen<br>
> in the layouted formula.<br>
><br>
> But that doesn't really matter when the formula is displayed since it<br>
> will work with any logically position calculated.<br>
> At most one needs to move all the formula nodes in order to have the<br>
> whole formula start at pos (0,0) if needs be...<br>
><br>
[..]<br>

Thanks Thomas for your very important notes. For archiving reasons (&
in case other folks around can help with ideas) I have taken whole of
reply into this mail and have cc to dev@openoffice and dev@education

Cheers!

--
rakesh

------------------------------------------

Hello Rakesh,

...
> >> 2. Now as per discussion which has happened so far, every Node needs<br>
> >> to have baseline calculated. To start with, I think I am ready to<br>
> >> write code for simple cases like "over". My strategy for over is<br>
> >> recursive calculation of baseline. After moving numerator to top and<br>
> >> denominator to bottom we will find the average of both and assign this<br>
> >> to baseline of whole fraction and all this will go on recursively<br>
> >> automatically. Does it sound good ?<br>
> >><br>
> >> something like(just a pseudo code ) :<br>
> >><br>
> >> if (pNum->HasBaseline() && pDenom->HasBaseline())<br>
> >> {<br>
> >> SetHasBaseline(TRUE);<br>
> >> SetBaseline((pNum->GetBaseline() +<br>
> >> pDenom->GetBaseline()) / 2L);<br>
> >> }<br>
> >><br>
> >> That needs fraction to be a node. And is it really a node ? I think yes.<br>
> >><br>
> >> These calculations are in SmBinVerNode::Arrange method.<br>
> >><br>
> ><br>
> > The basic idea is not quite right. If formulas are layouted it is not<br>
> > about finding the overall height and setting the mid as baseline.<br>
> > Layout of fractions should be done in away that the horizontal line is<br>
> > exactly on the same y position as a "-" in "a-b" would be if there were<br>
> > further formula parts.<br>
> ><br>
> > See for example: a - {a over b} over {y+z} or a - {H^{2^3}} over {H+H}<br>
> > And compare with any book about physics or math.<br>
> ><br>
><br>
> So, for simple cases like "<X> over <Y>" baseline would simply be<br>
> baseline of "over", if I am getting it right.<br>
><br>
That's not correct either. If the overline would have been drawn by
using the '' character that might have been right (exact result
depending on the font used, maybe). But in Math the underscore is drawn
as rectangle and thus is no character from any font. That means you have
to 'invent' a calculation for a pseudo baseline of that SmRect that fits
the requirement of aligning with the '-' character.

On the other hand you are free to re-implement that Node by making use
of the '-' character. You will need to call a separate function though
to stretch it to the desired width. And the layout may change somewhat
because of this.
However that will bring up other issue you need to take care of. For example
- Even if you stretch the '-' to the required width you need to make
sure that
the bounding rectangle does not include excessively large borders to
the left
and right.
- Also you need to clip the bounding rectangle at top and bottom to
about the
actual space the '-' will take because you must not have too much space
between it and the nominator and denominator.
- The advantage however would be that of course the baseline of '-' is
always
in line with another '-' as in "a - {x over y}". That is in that case the
baseline of the '-' should be the baseline of that node.
Note: if you want to go with the stretched '-' make sure to use the one
from the StarSymbol font. Not all fonts are arbitrarily scalable.

> >> 3. There are many messages and comments in French, and I would like to<br>
> >> start converting important ones as and when they come into English.<br>
> >> Would that be fine ? And can I do it in same CWS ?<br>
> >><br>
> > Err... Where are messages and comments in French?<br>
> > And do you want to put them into a CWS once you have them translated<br>
> > into English?<br>
> ><br>
><br>
> Actually they where German and few where in French (I am not sure<br>
> though, don't know either of them :( . I plan to start a page in wiki<br>
> and make a table with two columns. First one will contain messages in<br>
> non-English and other side will contain translated messges. For<br>
> critical comments related to this issue I will like to commit them in<br>
> this CWS.<br>
><br>
Sure. Do it that way.
Critical comment should best be in the source as well.

> Thanks Thomas for your very important notes. For archiving reasons (&<br>
> in case other folks around can help with ideas) I have taken whole of<br>
> reply into this mail and have cc to dev@openoffice and dev@education<br>
><br>
><br>
That's fine with me. :-)
Thanks for taking on that issue!

Thomas

-------------------------

Education Project/Effort/Math baseline alignment/Mails

2008-12-06T08:38:00Z

RakeshPandit:

-----------------------------------

Hello Thomas,

I had been debugging starmath module. Using some test cases and
checking the flow.

There are some very basic doubts:

1. In my view centre(0,0) of co-ordinate system from which all
calculations are made (e.g calculation of Point, Alignment etc) is at
top left corner. The +ve side of Y axis goes down the left corner
towards bottom and +ve side of X axis goes horizontal towards top
right corned. May you confirm this ?

2. Now as per discussion which has happened so far, every Node needs
to have baseline calculated. To start with, I think I am ready to
write code for simple cases like "over". My strategy for over is
recursive calculation of baseline. After moving numerator to top and
denominator to bottom we will find the average of both and assign this
to baseline of whole fraction and all this will go on recursively
automatically. Does it sound good ?

something like(just a pseudo code ) :

if (pNum->HasBaseline() && pDenom->HasBaseline())
{
SetHasBaseline(TRUE);
SetBaseline((pNum->GetBaseline() + pDenom->GetBaseline()) / 2L);
}

That needs fraction to be a node. And is it really a node ? I think yes.

These calculations are in SmBinVerNode::Arrange method.

3. There are many messages and comments in French, and I would like to
start converting important ones as and when they come into English.
Would that be fine ? And can I do it in same CWS ?

4. Should I reassign issue to myself and create a CWS? It hasn't been
created yet.

5. Can baseline be negative ? I don't think so, but there are test
cases where baseline has value -<number>

Thanks

--
Cheers,
rakesh

-----------------------------------

Hello Rakesh,

> Hello Thomas,
>
> I had been debugging starmath module. Using some test cases and
> checking the flow.
>
> There are some very basic doubts:
>
> 1. In my view centre(0,0) of co-ordinate system from which all
> calculations are made (e.g calculation of Point, Alignment etc) is at
> top left corner. The +ve side of Y axis goes down the left corner
> towards bottom and +ve side of X axis goes horizontal towards top
> right corned. May you confirm this ?
>
Yes. That is not what one might expect, but for all window based
coordinates in most UI API's it is the same.

> 2. Now as per discussion which has happened so far, every Node needs
> to have baseline calculated. To start with, I think I am ready to
> write code for simple cases like "over". My strategy for over is
> recursive calculation of baseline. After moving numerator to top and
> denominator to bottom we will find the average of both and assign this
> to baseline of whole fraction and all this will go on recursively
> automatically. Does it sound good ?
>
> something like(just a pseudo code ) :
>
> if (pNum->HasBaseline() && pDenom->HasBaseline())
> {
> SetHasBaseline(TRUE);
> SetBaseline((pNum->GetBaseline() +
> pDenom->GetBaseline()) / 2L);
> }
>
> That needs fraction to be a node. And is it really a node ? I think yes.
>
> These calculations are in SmBinVerNode::Arrange method.
>

The basic idea is not quite right. If formulas are layouted it is not
about finding the overall height and setting the mid as baseline.
Layout of fractions should be done in away that the horizontal line is
exactly on the same y position as a "-" in "a-b" would be if there were
further formula parts.

See for example: a - {a over b} over {y+z} or a - {H^{2^3}} over {H+H}
And compare with any book about physics or math.

> 3. There are many messages and comments in French, and I would like to
> start converting important ones as and when they come into English.
> Would that be fine ? And can I do it in same CWS ?
>
Err... Where are messages and comments in French?
And do you want to put them into a CWS once you have them translated
into English?

> 4. Should I reassign issue to myself and create a CWS? It hasn't been
> created yet.
>
If you feel like it you can do so.
That would be a good idea if you want to share the code with otehrs e.g.
Eric or me while developing.
If you want that just drop me a note and I will setup a CWS. That is
because in the end our QA will have a look at it and it is easier to get
a build done/installed from here if the CWS is on our servers...

> 5. Can baseline be negative ? I don't think so, but there are test
> cases where baseline has value -<number>
>
Yes it can become negative.
The actual baseline of the very outer line (or to put it different: of
the first char in a line) should not be negative. But it wouldn't matter.

The reason for baselines to become negative is because of aligning
different nodes vertically. If you take the following two formulas for
example
x - x
and
stack{a#b#c#d#e#f#g}
each on it's own none of the baselines should be negative.
But if they are on the same line like
x - x stack{a#b#c#d#e#f#g}
things may become different.
The reason for this is that the alignment of the whole formula starts
with the aligned node. And here that is 'x'. If, just for the example,
we assume a lines height will be 100 then the rectangle for 'a' might be
at pos 0,0 with size 60x100.
And the stack if placed at (0,0) would have a size of about 160x700.
If the latter one is to be vertically aligned to the first it is done by
moving it up until in order for the baseline of the stack{...} command
to match the one from "x-y". By doing that we surely will have the
baselines of a, b and c be moved into the negative y axis as can be seen
in the layouted formula.

But that doesn't really matter when the formula is displayed since it
will work with any logically position calculated.
At most one needs to move all the formula nodes in order to have the
whole formula start at pos (0,0) if needs be...

Thomas

-------------------------------------

2008/9/29 Thomas Lange - Sun Germany - ham02 - Hamburg <Thomas.Lange@sun.com>:
[..]
- Show quoted text -
>> There are some very basic doubts:
>>
>> 1. In my view centre(0,0) of co-ordinate system from which all
>> calculations are made (e.g calculation of Point, Alignment etc) is at
>> top left corner. The +ve side of Y axis goes down the left corner
>> towards bottom and +ve side of X axis goes horizontal towards top
>> right corned. May you confirm this ?
>>
> Yes. That is not what one might expect, but for all window based
> coordinates in most UI API's it is the same.
>
>> 2. Now as per discussion which has happened so far, every Node needs
>> to have baseline calculated. To start with, I think I am ready to
>> write code for simple cases like "over". My strategy for over is
>> recursive calculation of baseline. After moving numerator to top and
>> denominator to bottom we will find the average of both and assign this
>> to baseline of whole fraction and all this will go on recursively
>> automatically. Does it sound good ?
>>
>> something like(just a pseudo code ) :
>>
>> if (pNum->HasBaseline() && pDenom->HasBaseline())
>> {
>> SetHasBaseline(TRUE);
>> SetBaseline((pNum->GetBaseline() +
>> pDenom->GetBaseline()) / 2L);
>> }
>>
>> That needs fraction to be a node. And is it really a node ? I think yes.
>>
>> These calculations are in SmBinVerNode::Arrange method.
>>
>
> The basic idea is not quite right. If formulas are layouted it is not
> about finding the overall height and setting the mid as baseline.
> Layout of fractions should be done in away that the horizontal line is
> exactly on the same y position as a "-" in "a-b" would be if there were
> further formula parts.
>
> See for example: a - {a over b} over {y+z} or a - {H^{2^3}} over {H+H}
> And compare with any book about physics or math.
>

So, for simple cases like "<X> over <Y>" baseline would simply be
baseline of "over", if I am getting it right.

>> 3. There are many messages and comments in French, and I would like to
>> start converting important ones as and when they come into English.
>> Would that be fine ? And can I do it in same CWS ?
>>
> Err... Where are messages and comments in French?
> And do you want to put them into a CWS once you have them translated
> into English?
>

Actually they where German and few where in French (I am not sure
though, don't know either of them :( . I plan to start a page in wiki
and make a table with two columns. First one will contain messages in
non-English and other side will contain translated messges. For
critical comments related to this issue I will like to commit them in
this CWS.

>> 4. Should I reassign issue to myself and create a CWS? It hasn't been
>> created yet.
>>
> If you feel like it you can do so.
> That would be a good idea if you want to share the code with otehrs e.g.
> Eric or me while developing.
> If you want that just drop me a note and I will setup a CWS. That is
> because in the end our QA will have a look at it and it is easier to get
> a build done/installed from here if the CWS is on our servers...
>

Okay.
I will do it, thanks
- Show quoted text -

>> 5. Can baseline be negative ? I don't think so, but there are test
>> cases where baseline has value -<number>
>>
> Yes it can become negative.
> The actual baseline of the very outer line (or to put it different: of
> the first char in a line) should not be negative. But it wouldn't matter.
>
> The reason for baselines to become negative is because of aligning
> different nodes vertically. If you take the following two formulas for
> example
> x - x
> and
> stack{a#b#c#d#e#f#g}
> each on it's own none of the baselines should be negative.
> But if they are on the same line like
> x - x stack{a#b#c#d#e#f#g}
> things may become different.
> The reason for this is that the alignment of the whole formula starts
> with the aligned node. And here that is 'x'. If, just for the example,
> we assume a lines height will be 100 then the rectangle for 'a' might be
> at pos 0,0 with size 60x100.
> And the stack if placed at (0,0) would have a size of about 160x700.
> If the latter one is to be vertically aligned to the first it is done by
> moving it up until in order for the baseline of the stack{...} command
> to match the one from "x-y". By doing that we surely will have the
> baselines of a, b and c be moved into the negative y axis as can be seen
> in the layouted formula.
>
> But that doesn't really matter when the formula is displayed since it
> will work with any logically position calculated.
> At most one needs to move all the formula nodes in order to have the
> whole formula start at pos (0,0) if needs be...
>
[..]

Thanks Thomas for your very important notes. For archiving reasons (&
in case other folks around can help with ideas) I have taken whole of
reply into this mail and have cc to dev@openoffice and dev@education

Cheers!

--
rakesh

------------------------------------------

Hello Rakesh,

...
> >> 2. Now as per discussion which has happened so far, every Node needs
> >> to have baseline calculated. To start with, I think I am ready to
> >> write code for simple cases like "over". My strategy for over is
> >> recursive calculation of baseline. After moving numerator to top and
> >> denominator to bottom we will find the average of both and assign this
> >> to baseline of whole fraction and all this will go on recursively
> >> automatically. Does it sound good ?
> >>
> >> something like(just a pseudo code ) :
> >>
> >> if (pNum->HasBaseline() && pDenom->HasBaseline())
> >> {
> >> SetHasBaseline(TRUE);
> >> SetBaseline((pNum->GetBaseline() +
> >> pDenom->GetBaseline()) / 2L);
> >> }
> >>
> >> That needs fraction to be a node. And is it really a node ? I think yes.
> >>
> >> These calculations are in SmBinVerNode::Arrange method.
> >>
> >
> > The basic idea is not quite right. If formulas are layouted it is not
> > about finding the overall height and setting the mid as baseline.
> > Layout of fractions should be done in away that the horizontal line is
> > exactly on the same y position as a "-" in "a-b" would be if there were
> > further formula parts.
> >
> > See for example: a - {a over b} over {y+z} or a - {H^{2^3}} over {H+H}
> > And compare with any book about physics or math.
> >
>
> So, for simple cases like "<X> over <Y>" baseline would simply be
> baseline of "over", if I am getting it right.
>
That's not correct either. If the overline would have been drawn by
using the '' character that might have been right (exact result
depending on the font used, maybe). But in Math the underscore is drawn
as rectangle and thus is no character from any font. That means you have
to 'invent' a calculation for a pseudo baseline of that SmRect that fits
the requirement of aligning with the '-' character.

On the other hand you are free to re-implement that Node by making use
of the '-' character. You will need to call a separate function though
to stretch it to the desired width. And the layout may change somewhat
because of this.
However that will bring up other issue you need to take care of. For example
- Even if you stretch the '-' to the required width you need to make
sure that
the bounding rectangle does not include excessively large borders to
the left
and right.
- Also you need to clip the bounding rectangle at top and bottom to
about the
actual space the '-' will take because you must not have too much space
between it and the nominator and denominator.
- The advantage however would be that of course the baseline of '-' is
always
in line with another '-' as in "a - {x over y}". That is in that case the
baseline of the '-' should be the baseline of that node.
Note: if you want to go with the stretched '-' make sure to use the one
from the StarSymbol font. Not all fonts are arbitrarily scalable.

> >> 3. There are many messages and comments in French, and I would like to
> >> start converting important ones as and when they come into English.
> >> Would that be fine ? And can I do it in same CWS ?
> >>
> > Err... Where are messages and comments in French?
> > And do you want to put them into a CWS once you have them translated
> > into English?
> >
>
> Actually they where German and few where in French (I am not sure
> though, don't know either of them :( . I plan to start a page in wiki
> and make a table with two columns. First one will contain messages in
> non-English and other side will contain translated messges. For
> critical comments related to this issue I will like to commit them in
> this CWS.
>
Sure. Do it that way.
Critical comment should best be in the source as well.

> Thanks Thomas for your very important notes. For archiving reasons (&
> in case other folks around can help with ideas) I have taken whole of
> reply into this mail and have cc to dev@openoffice and dev@education
>
>
That's fine with me. :-)
Thanks for taking on that issue!

Thomas

-------------------------

Education Project/Effort

2008-06-06T16:15:44Z

RakeshPandit:

{{:Uno/Term/Effort}}

We propose '''applications''', doable in one or two month of work for a student ( time to discover everything, and do the real work).
All these applications below are correlated with '''IssueZilla issues''', and need new code to be solved.

What we propose is to accompany the peers teachers/students solving the issues: the work must be done by the student, and we are resources. This means, most of the time, we have no clue about the solution (but we could work on it), or the needed changes, and sometimes, just a little change in the code can solve the issue. Sometimes, a lot of code can be needed.

The most important is to work together, and make the students discover OpenOffice.org source code and organisation.

The applications are classed

- by difficulty level (more stars means means more work and more difficult to be solved )

- by themes. e.g. Mac OS X

Last but not least, have a look at [[Cpp_Coding_Standards]] cannot be bad :-)

== '''Applications for Mac OS X only''' ==

=== [[Education_Project/Effort/Write_a_new_canvas | Implement new canvas using Core Graphics (click me)]]===

FIXME: description needs to be updated

* '''Goal''' : Implement a new canvas on Mac OS X

* '''Problem''' : nothing yet

* '''Time plan''' : TBD

Contacts:

* for Education project: ericb , (please add your name )

* for the task : Thorsten Behrens thb/ Ismael Merzaq ismael87 / Eric Bachard ericb : all @openofice.org

Difficulty: *****

Needs:

* experimental tests in canvas, with dummy methods, for proof of concept

* good Core Graphics understanding

* good C++ skills

* good Cairo understanding ( we use the existing implementation)

=== [[Education_Project/Effort/Complete_feature_send_active_document_as | Support of more mail clients for send active document as (click me)]] ===

* '''Goal''' : Support multiple mail client applications on Mac OS X (currently Mail.app only) and determine the user's default mail client for the "send document as e-mail" feature

* '''Problem''' : Only Mail.app support creating a new message with a document already attached using the 'open -a' command.

* '''Time plan''' : TBD

Contacts:

* for Education project: ericb , (please add your name )
* for the task : obr @openofice.org

Difficulty: ***

=== [[Education_Project/Effort/Use_system_security_API | Use system security API ( click me)]] ===

* '''Goal''' : Make Apple key signing API work

* '''Problem''' : currently, a strange hack does use Mozilla libnss and co, and this is not easy to maintain. Use Apple API would simplify the build, and help to improve the OpenOffice.org integration on Mac OS X.

* '''Time plan''' : long term work. TBD completely

* '''Links''' : [http://developer.apple.com/documentation/Security/Conceptual/keychainServConcepts/01introduction/chapter_1_section_1.html Apple Keychain Documentation]

Contacts:

* for Education project: ericb , (please add your name )
* for the task : jl/obr @openofice.org

Difficulty: *****

* no information available yet

=== [[Education_Project/Effort/Improve_Scanner_Use | Improve Scanner Use (click me)]] ===

==== '''Description of the application''' ====

* '''Goal''' : Implement Apple way of Image Capture

* '''Target''' : Mac OS X Aqua version is concerned firstly (not sure a backport is possible)

* '''Problem''' : the current implementation :

- needs a third part installation ( libsane , libusb ..etc )

- is obsolete

- is not User Experience compliant

* '''Strategy (draft, to be completed)'''

Mac OS X does use a blob interfaced with most of existing scanners (undocumented), named " Image Capture ". This blob seems to use Twain implementation.

The scanner is an abstraction in every subpart like scandlg.cxx grid.cxx, and scnserv.cxx. '''The Mac OS X implementation does only concern scanunx.cxx'''.

The problem is, scanunx.cxx does use SANE implementation, and will become useless. Thus, '''the idea is to create scanmacosx.cxx, and use both scanwin.cxx existing implementation . the result will be a mix of what Image Catpure SDK + Apple documentation will provide us, and the need we have for the implementation'''.

We firstly have to :

- analyze scanunx.cxx completely

- design TwainHolder, TwainThread, BitmapManager, ScannerManager (reused from scanwin.cxx ) implementation in scanmacosx.cxx

- define a process for the progressive implementation

- FIXME: complete ..

* '''Time plan''' : a first try must work on early of June 2008

===='''Contacts''' ====

* Volunteers :

- '''Valentin JANIAUT''' student from [http://www.utbm.fr '''UTBM (France )''']

- [[User:Dyrcona | '''Jason STEPHENSON''']], from USA

- [[User:Ericb | '''Eric Bachard''']]

Jason is a skilled developer ( Cocoa ) helping for the feature implementation.

Valentin should be able to validate a Value Unit for his own cursus, writing code for OpenOffice.org project.

* Contact for School : Eric Bachard

* for Education Project : ericb , (please add your name )

* To be defined

This application is a try. Main idea is : analyse current interface, improve the current implementation, and more if possible.

Difficulty: ****

Needs:

* C++ skills

* Twain and Sane skills

* interface understanding (offapi )

* Design skills

== '''Application concerning all ports''' ==

=== [[Education_Project/Effort/Math_baseline_alignment | Implement Math Baseline Alignement (click me)]] ===

* '''Goal''' : Fix [http://www.openoffice.org/issues/show_bug.cgi?id=972 issue972] Implementing a correct baseline alignment for equations anchored as characters in Writer documents ( should work for "to character" too ).

Side effects: a change in .odt specification will result, and specs must be written in parallel

* '''Problem''' : the current implementation does not work, and equations are misaligned.

'''Important consequence: OpenOffice.org acceptance is low in Educationnal system (all locales concerned)'''

* '''Time plan''' : (will be defined soon)

Note: the application could be proposed as Google Summer of Code ( including other tasks.

Contacts:

* for Education Project : ericb , (please add your name )
* Framework/Starmath projects : mba/tl @openoffice.org

Difficulty: ****

New: needs

* new interface definition ( offapi),
* new file format extension ( needs ODF TC agreement ),
* new implementation ( starmath )
* Serious QA

Skills : C++

=== [[Education_Project/Effort/Uno-skeletonmaker templates | Make the uno-skeletonmaker tool templatable (click me)]] ===

* '''Goal''' : Modify or rewrite the uno-skeletonmaker to make it templatable.

* '''Problem''' : the current implementation is completely hard-coded. Every developer has its own development style guide and has to change the generated code every time.

* '''Time plan''' : (will be defined soon)

Note: the application could be proposed as Google Summer of Code ( including other tasks).

Contacts:

* for Education Project : ericb , cedricbosdo (for development)

Difficulty:

New: needs

* understanding the existing uno-skeletonmaker (unodevtools)

Skills : C++

=== [[Education_Project/Effort/Fix_Mousewheel_Jump_in_Calc | Fix MouseWheel jump in Calc (click me)]] ===

==== '''Description of the application''' ====

* '''Goal''' : Fix [http://www.openoffice.org/issues/show_bug.cgi?id=81907 issue 81907] Implementing a method allowing smooth scrolling in Calc sheets, even when cells are half if the screen high

* '''Problem''' : the current implementation uses integers for the scrolling, and whenever a cell is half of the screen high, some cells are no longer accessible.

* '''Time plan''' : (will be defined soon)
Contacts:

* for Education Project : ericb , (please add your name )

* Framework Project : nn/ @openoffice.org

Difficulty: **

Needs:

* a correct understanding of the existing code

* define the new feature fixing the issue

* implement it

* integrate it respecting OpenOffice.org project rules

=== Create a special version for Educational purpose ===

Points to be improved :

* implement larger icons
* remove zillions of useless menus
* remove base and calc libs
* remove useless stuff

==== [[Education_Project/Effort/Registration Screen | Registration Screen]]====

* '''Goal''' : allow to remove the registration screen for e.g. school or entities where the features becomes evil

* '''Problem''' : see [http://www.openoffice.org/issues/show_bug.cgi?id=89170 '''Issue 89170'''] Currently not implemented.

* '''Time plan''' : TBD

Contacts:

* for Education project: ericb , (please add your name )

Difficulty: No estimation yet

Concerned Domains : Framework, User Experience

=== [[Education_Project/Effort/Fixing_warning | Fixing warnings (click me)]] ===
FIXME: needs to be updated

* '''Goal''' : Fix all warnings coming from old modules as well as fix bugs

* '''Problem''' : building any module prints various warnings on stderr, ideally we need zero warnings.

* '''Time plan''' : TBD

Contacts:

* for education project: ericb, rakeshpandit (rakesh.pandit@gmail.com)

Difficulty:

* Easy

Needs:

* C++

Comments:

Before starting debugging you can coordinate by updating the files you are debugging in
[[Education_Project/Effort/Fixing_warning | here]].

=== [[Education_Project/Effort/Integrate_scim | Integrate SCIM into OO (click me)]]===

FIXME: description needs to be updated

* '''Goal''' : Integrate SCIM(Smart Common Input Method) into OO.

* '''Problem''' : Currently not implemented. As an example, this is available in Gedit and Abiword.

* '''Time plan''' : TBD

Contacts:

* for Education project: ericb , (please add your name )

Difficulty:

* http://www.scim-im.org/wiki/faq/gtk_gnome/why_i_can_not_input_with_scim_under_openoffice_oo

Needs:

Comments:
This is not an OpenOffice.org related topic. This is a problem of the distribution configuration. For example on Ubuntu, read this page and it works:
http://doc.ubuntu-fr.org/scim.
[[User:Cedricbosdo|cedricbosdo]] 18:17, 11 April 2008 (CEST)

== '''Other Applications (later)''' ==

===[[Education_Project/Effort/Maxima Integration |Maxima Integration (click me)]] ===



=== [[Education_Project/Effort/OpenOffice.org_Installation | Write OpenOffice.org Installation HowTo]] ===

[[Category:Education]]

[[Category:Effort]]

Education Project/Effort/Fixing warning

2008-06-06T15:46:29Z

RakeshPandit: /* Modules and respective files which are currently being debugged. */

Education Project/Effort/Fixing warning

2008-06-06T15:45:43Z

RakeshPandit: New page: === Modules and respective files which are currently being debugged. === Please put in dates when entered and expected time in which you will release lock, so that in case someone forgets...

Education ClassRoom/Agenda

2008-05-18T04:56:39Z

RakeshPandit:

'''IMPORTANT: this is a draft, and some developers mentioned below, didn't confirm yet they are ok to attend.'''

{| style="vertical-align:top; text-align:left; background-color:#efefef;"
|- style="background-color:#d6d6d6;font-weight:bold;"
| Topic || Guest || Date
|-
| OpenOffice.org source code scheduling || XXX || undefined (no confirmation)
|-
| FrameWork Project || Mathias Bauer || June 4th
|-
| Extensions Project || Jürgen Schmidt ? || canceled (no confirmation)
|-
| GSL Project || Philipp Lohmann || 21st of May
|-
| UDK and UNO projects || undefined || undefined
|-
| Tinderboxes System || Christian Lohmaier|| May 19
|-
| Calc || Eike Rathke || ok, but after 3.0
|}

'''Prefered hour is 11:00 ( Paris /Hambourg hour)''' because it is more easy for Indian students/attendees .

{| style="vertical-align:top; text-align:left; background-color:#efefef;"
|- style="background-color:#d6d6d6;font-weight:bold;"
| Date ||Who || Subject || Targeted audience || Links ||
|-
|-bgcolor="#a6ffa4" |
| '''May 7th''' || '''Eric Bachard''' || '''Discover OpenOffice.org Project: the development side''' || Students and potential devs who don't know how start ||
|-
| '''May 19th 11:00-12:00''' || '''Christian Lohmaier''' || '''All about Tinderboxes''' || all interested in automatic building/what to do with the results/why tinderboxes are a good thing to have :-)
|-
|-bgcolor="#a6ffa4" |
| '''May 21st''' || '''Philipp Lohmann''' || '''Graphic System Layer Project''' || C++ coders || [http://gsl.openoffice.org/ GSL Project]
|-
| May 28th to 30th
|-
|-bgcolor="#a6ffa4" |
| '''June 4th''' || '''Mathias Bauer''' || '''Framework Project''' || C++ developers that want to work on the core source code || [http://framework.openoffice.org Framework Project]
|-
| June 11th to 13th
|-
| June 18 to 20th
|}

[[Category:Education]]

[[Category:Project]]

[[Category:Projects]]