R help - Aug 2008 - Possible buglet (wart ?) of odfWeave 0.7.6 (with workaround)

Dear List,

I have had problems inserting some (not all !) figures via odfWeave
(using print(someLatticeFunction)...). The figure was correctly
displayed in a R device window but the resulting ODF document displayed
the correct space for the figure and an empty frame with a "broken
image" icon and a "read error" mention.

Exploration of the odf (.odt, in my case) file showed that the relevant
files (.eps, which is a vector format much more apt than bitmaps to
cutting/pasting on various media) existed in the Pictures directory of
the file and displayed fine.

Then lightning struck : the picture files were named an follows :
"content_",  serial number, chunk name, ".eps" ; it occured
to me that
OpenOffice 2.4.1 (as packaged in Debian) displayed a "Read error"
*when
the chunk name had accented characters in it* (my system works in UTF8).
Plain ASCII chunk names work OK.

I do not know ODF specification well enough to be sure that this is an
odfWeave problem (some chunk name mangling might be necessary) rather
than an OpenOffice "infelicity" (non-implementation of a
specification-requested feature) : checking with another ODF-compliant
word processor might be useful, but I don't have any on hand.

In any case, this problem is not obvious after re-reading the odfWeave
documentation, and does not seem to have been already reported to r-help
(yes, Pr Ripley, I did my homework...). This report might spare some
poor non-English-writer sod some time ...

For what it's worth :
> sessionInfo()
R version 2.7.1 (2008-06-23) 
x86_64-pc-linux-gnu 

locale:
LC_CTYPE=fr_FR.UTF-8;LC_NUMERIC=C;LC_TIME=fr_FR.UTF-8;LC_COLLATE=fr_FR.UTF-8;LC_MONETARY=C;LC_MESSAGES=fr_FR.UTF-8;LC_PAPER=fr_FR.UTF-8;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=fr_FR.UTF-8;LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods
base     

other attached packages:
[1] odfWeave_0.7.6  XML_1.96-0      lattice_0.17-13

loaded via a namespace (and not attached):
[1] grid_2.7.1  tools_2.7.1

Hoping this might help,

					Emmanuel Charpentier

PS : don't try to answer to my subscribed address (charpent (at) bacbuc
(dot) dyndns (dot) org), which is the apparent source of this message :
bacbuc (dot) dyndns (dot) org is down and will quite like likely
remain so for at least two to three more weeks. I'm playing fast and
loose with Gmane by putting my registered address as a Reply-to address
and use my actual present address (emm (dot) charpentier (at) free (dot) fr)
as source. I do not know what will survive Gmane, Mailman and Martin's
scripts
mauling...

Dear List,

I have had problems inserting some (not all !) figures via odfWeave
(using print(someLatticeFunction)...). The figure was correctly
displayed in a R device window but the resulting ODF document displayed
the correct space for the figure and an empty frame with a "broken
image" icon and a "read error" mention.

Exploration of the odf (.odt, in my case) file showed that the relevant
files (.eps, which is a vector format much more apt than bitmaps to
cutting/pasting on various media) existed in the Pictures directory of
the file and displayed fine.

Then lightning struck : the picture files were named an follows :
"content_",  serial number, chunk name, ".eps" ; it occured
to me that
OpenOffice 2.4.1 (as packaged in Debian) displayed a "Read error"
*when
the chunk name had accented characters in it* (my system works in UTF8).
Plain ASCII chunk names work OK.

I do not know ODF specification well enough to be sure that this is an
odfWeave problem (some chunk name mangling might be necessary) rather
than an OpenOffice "infelicity" (non-implementation of a
specification-requested feature) : checking with another ODF-compliant
word processor might be useful, but I don't have any on hand.

In any case, this problem is not obvious after re-reading the odfWeave
documentation, and does not seem to have been already reported to r-help
(yes, Pr Ripley, I did my homework...). This report might spare some
poor non-English-writer sod some time ...

For what it's worth :
> sessionInfo()
R version 2.7.1 (2008-06-23) 
x86_64-pc-linux-gnu 

locale:
LC_CTYPE=fr_FR.UTF-8;LC_NUMERIC=C;LC_TIME=fr_FR.UTF-8;LC_COLLATE=fr_FR.UTF-8;LC_MONETARY=C;LC_MESSAGES=fr_FR.UTF-8;LC_PAPER=fr_FR.UTF-8;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=fr_FR.UTF-8;LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods
base     

other attached packages:
[1] odfWeave_0.7.6  XML_1.96-0      lattice_0.17-13

loaded via a namespace (and not attached):
[1] grid_2.7.1  tools_2.7.1

Hoping this might help,

					Emmanuel Charpentier

PS : don't try to answer to my subscribed address (charpent (at) bacbuc
(dot) dyndns (dot) org), which is the apparent source of this message :
bacbuc (dot) dyndns (dot) org is down and will quite like likely
remain so for at least two to three more weeks. I'm playing fast and
loose with Gmane by putting my registered address as a Reply-to address
and use my actual present address (emm (dot) charpentier (at) free (dot) fr)
as source. I do not know what will survive Gmane, Mailman and Martin's
scripts
mauling...

Emmanuel,

On 8/12/08 1:13 PM, "Emmanuel Charpentier" <emm.charpentier at
free.fr> wrote:
 
> Then lightning struck : the picture files were named an follows :
> "content_",  serial number, chunk name, ".eps" ; it
occured to me that
> OpenOffice 2.4.1 (as packaged in Debian) displayed a "Read error"
*when
> the chunk name had accented characters in it* (my system works in UTF8).
> Plain ASCII chunk names work OK.
Since we use Sweave directly and that is the naming convention that it uses
for image files. Once you open the document in OpenOffice, it sometimes
renames the image files that are stored internally (in some versions of OO).
> I do not know ODF specification well enough to be sure that this is an
> odfWeave problem (some chunk name mangling might be necessary) rather
> than an OpenOffice "infelicity" (non-implementation of a
> specification-requested feature) : checking with another ODF-compliant
> word processor might be useful, but I don't have any on hand.
 
I think that this is an OpenOffice issue. Look at the last code chunk in:

demoFile <- system.file(
                        "examples",
                        "testCases.odt",
                        package ="odfWeave")

The code chunk is entirely Chinese characters. Apart from the current
encoding issues with odfWeave and XML (on Windows only), this code chunk
works fine. If I add fig = TRUE and generate a plot, I get the same issue
you described (using NeoOffice and OpenOffice). odfWeave can deal with these
characters in code and in code chunk names.

I agree that it *should* work since XML documents should be fully UTF-8
compatible. I did a quick search of there open bugs, but I didn't see
anything. You should document the issue and submit a bug report.


Max