mdb-export prints binary for OLE fields.

[Adam Buchbinder]

Binary package hint: mdbtools

If the database contains OLE fields, mdb-export will output these as raw binary, causing character encoding issues, at the very least, and in my case, unreadable data--the fields are apparently used for replication; this one is "s_Lineage". Due to some unrelated problems with Access, I can't remove the fields, so I'm stuck trying to deal with this input data.

A simple fix is to escape the data in OLE fields as octal bytes, e.g., \130 or whatnot. This requires a rather localized change; a patch to do so (applying to mdb-export.c) is attached. The resultant CSV files now validate with csv-check from the libtext-csv-xs-perl examples.

Unfortunately, I can't attach the database; I've only been able to reproduce the problem with a copy of our production data, which (a) is over 100M, and (b) I'm not privileged to post. But if anyone wants to attach a database with OLE fields that exports mangled data with the unpatched version, please feel free.

ProblemType: Bug
Architecture: i386
DistroRelease: Ubuntu 9.04
Package: mdbtools 0.5.99.0.6pre1.0.20051109-4
ProcEnviron:
 PATH=(custom, user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: mdbtools
Uname: Linux 2.6.28-11-generic i686

[Adam Buchbinder (adam-buchbinder)]

I should note that the current patch mangles the fields, e.g., I get '\147\37777777767\2' where I should probably be getting something else. It doesn't output broken SQL (well, embedded newlines in text fields aren't escaped, but that's a different problem), so it at least works for me; I have no need for these fields to be usable.

[Jean-Michel Vourgère (jmv-deb)]

Actually this is a bit more complex.
When using csv format, one should *not* escape the binary.
When using insert statement, characters should be escaped in a backend specific way...

[Jean-Michel Vourgère (jmv-deb)]

Fixed upstream:
https://github.com/brianb/mdbtools/commit/a66531c74a7a28cf91c8699f12a0182429f47972