This script and plug-in checks all the URL fields of a repository and updates the issue list. It does so for daily batches of eprint items or for all items. The field names are taken from a configuration file. Upon request, it also produces a report of the checked URLs sorted by HTTP status code.
Copy the contents of the bin and lib directories to the respective {eprints_root}/bin and {eprints_root}/lib directories. Copy the cfg/cfg.d/z_linkcheck.pl file to {eprints_root}/archives/{yourarchive}/cfg/cfg.d/ z_linkcheck.pl. Optionally, if you want to have improved help for Admin > Issues search, copy the phrases files cfg/lang/{language}/phrases/eprint_item_issues.xml to their respective {eprints_root}/archives/{yourarchive}/cfg/lang/{language}/phrases/eprint_item_issues.xml.
Edit the {eprints_root}/archives/{yourarchive}/cfg/cfg.d/z_linkcheck.pl file according to your requirements.
Restart the web server.
Test your setup by checking an eprint that has an invalid, a working or a non-functioning URL:
sudo -u apache {eprints_root}/bin/linkcheck {repo} {eprintid} --verbose --report
LinkCheck updates your item issues table with a issue type check_url_status for
non-functioning URLs (format-valid URLs that yield a HTTP status code <> 200) or
invalid_url for URLs that have an invalid format (e.g. a typo into the protocol,
such as htpp).
Detected issues are marked with issue status reported.
Issues where URLs were fixed (either removed or corrected) are marked with issue status
resolved upon subsequent run of linkcheck.
To find issues, login to your repository as administrator. Goto Admin > Search Issues,
and search by check_url_status or invalid_url in the Issue Type field.
However, we also recommend to use the --report option of linkcheck that produces a neat
report sorted by HTTP status code, eprintid and URL, because
- Search Issues does not provide a nice overview of the issues because it reports by eprint item, not by issue; there should be an option to group issues by issue description.
- there is no way to reset an issue manually after an URL has been corrected (only by running linkcheck again)
The report sorted by URL allows to find patterns of URLs that can be fixed by a batch procedure.
Depending on the number of URLs you have in your repository, you may either carry out nightly runs (e.g. as a cron job), which does a daily segment of your repository
linkcheck {repo} --report >{eprints_root}/var/linkcheck.txt
or a full run using --all
linkcheck {repo} --all --report >{eprints_root}/var/linkcheck.txt
A full run may take a long, long time (days).
Instead of using the separate linkcheck script, the issues audit procedure could be
invoked (see https://wiki.eprints.org/w/Issues_audit and
https://wiki.eprints.org/w/Using_Issues_for_Quality_Control ). Issues audit is usually
run nightly and processes all eprint item; however, depending on size of a repository and
the quality of the URLs, a link check may not finish in good time. We have thought about
this option, but decided to create a script that can be run separately and can process
URLs in smaller chunks.
For small repositories, issues audit might be an alternative and in some time we may
provide an issues plugin for URL checks.