https://bz.apache.org/ooo/show_bug.cgi?id=45593

--- Comment #22 from orcmid <[email protected]> ---
I did some mining in the ODF 1.2 specification to see if there are any
limitations there.  Basically, the answer is "no."

You can see the main schema definitions for length-related values at
http://nfoworks.org/notes/2014/05/n140504f.htm#length

The specification says the units of measure are as defined in [XSL] here:
http://www.w3.org/TR/2001/REC-xsl-20011015/slice5.html#section-N8185-Definitions-of-Units-of-Measure

However, the "em" unit is not defined in the ODF 1.2 schema for the various
length cases.  Instead, there is a separate relative length that is a no-unit
integer, so that is evidently not the equivalent of an "em", which can be
fractional.

FIRST CHECKS

It would be useful to determine whether this problem is part of the input
conversion when lengths are specified, and not actually a limitation on what is
acceptable in the file format itself.  The format allows arbitrary decimal
fixed-point values.  

It would be good to generate some test documents where the stored values have
more decimal precision to see what happens when those documents are ingested by
AOO (and other implementations of ODF).  If that succeeds, it confirms that the
limitation is from the input-conversion when lengths are being specified. 
That's the ideal case.  If there is truncation when reading the format, perhaps
as part of conversion to internal units, that's more serious.

It is also important to check whether the conversion of 0.375 to 0.38 is simply
in output conversion of the internal representation.  I assume that the
differences are apparent and the rounded result is what is actually used, based
on the reports here.

TRANSFORMATION ISSUES

Because graphics allow transformation of coordinate systems, there is a
requirement for greater precision in the internal processing.  @alg has
mentioned a concern about that in a previous comment.  Whatever is done to
carry more-precise length values has to carry into coordinate transformations.

INTEROPERABILITY CONSIDERATIONS

This defect (I agree, it is a bug in implementation-dependent behavior) is
inherited in openoffice.org common code.  The analysis of the defect and what
the repairs are needs to be shared with other descendants of openoffice.org. 
It would also be interesting to see the FIRST CHECKS results produced by
Microsoft Office Word when saving in ODF format.

-- 
You are receiving this mail because:
You are on the CC list for the issue.
You are the assignee for the issue.

Reply via email to