Hi Andreas, Sorry, the data was actually rotten for multiple reasons. First, the machine running upstream-metadata.debian.net stopped keeping dep-src entries in its sources.list, so debcheckout was not working anymore, and my rudimentary scripts did not catch the error. I added error-catching to the TODO list. Second, when packages change their repository URL, which is not supposed to happen often, they have to be refreshed by hand. Third, I hardcoded the erroneous git url git://git.debian.org/git that we now correct in git://git.debian.org/.
I have reloaded the data from scratch, by deleting the database and running the following command for each package med-bio depends on. curl http://upstream-metadata.debian.net/$package/YAML-URL I am now injecting all the fields related to bibliography. By the way, I regret that I have put PMID and DOI outside the Reference-* namespace. Would you mind if I correct this ? The file used for injection, http://upstream-metadata.debian.net/for_UDD/biblio.yaml, is valid YAML; this is why I managed to write the loader. It is a serie of records, which all contain an array of three fields. Altogether, they are loaded as a table of three columns. upstream-metadata.debian.net stores its data in a Berkeley database, where the field names are the concatenation of the package name and the upstream-metadata.yaml field name, that is, if in the perlprimer package, there is “PMID: 15073005”, the Berkeley DB will contain “15073005” for the field “perlprimer:PMID”. In the whole information chain, the structure is always ‘package - field - value’. I do not know where the perlprimer duplicaion came. Perhaps there was an invisible character somewhere ? On the server side, there is a command line tool to manipulate field values directly, I may have done a typo when making tests. This said, I agree that the output should be sanitized. Also, I definitely agree to use PRIMARY KEY (package,key) as an extra safety net. Should it be added to udd/sql/bibref.sql ? Cheers, -- Charles -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/[email protected]
