Links to dynamic pages

[James-Crean]

Concerning this page: https://pcmdi.llnl.gov/ipcc/subproject_publications.html

I believe this is the largest source of broken links on the site. There are about 1200 rows in this html table and almost all of them point toward 1 of 2 dynamic pages.

For reference here are two example pages on the old site:

• https://www-pcmdi.llnl.gov/ipcc/abstract.php?ipcc_publication_id=466&r=subprojects
• https://www-pcmdi.llnl.gov/ipcc/project_detail.php?ipcc_subproject_id=1146

Due to the fact that there is no publication information on the page, I cant just do a find and replace. All of the information about the publications is held in a database.

@durack1, @taylor13 Thoughts?

[taylor13]

Ideally, we would harvest the doi's (or obtain them from the citation information) reached via the pub1, pub2, .... pubN links that appear at the right side of the subproject_publications page. We would then use these to populate a new "project" (CMIP3) that would get added to https://cmip-publications.llnl.gov/search?type=project . That way users could access the CMIP3 publications in the same way they access CMIP5 publications.

How much time would it take (do you think) to extract all the actual publications from the CMIP3 database, then fill in the missing information (e.g., using various online services which Paul and I can point you to) and place the completed information into the new-style publication database used for CMIP5?

If this were not too difficult, we would then simply remove all the links from https://pcmdi.llnl.gov/ipcc/subproject_publications.html (which we probably will do in any case), and also remove the the right column with pub1, pub2, etc. from the right side of that page, leaving a simple listing of subprojects (without any links). We would add a sentence at the top of the page directing those interested to the "publication" listing for CMIP3 (if we in fact decide to do this.)

Let us know what you think.

[James-Crean]

It shouldn't take too long to get a list of DOI's. I would find and copy the database down to my local machine and create a script to pull out the DOI info from each publication assuming it had one. For any publications that lack a DOI, I would save the relevant information such as the title, and those would likely just need to be run through google to find a DOI.

By my approximation, I believe there could be as many as 600 publications that would need to be entered. This process is not particularly difficult, but could be rather time consuming as the site is not designed for mass entry of publications. They must be entered one at a time.

I am not sure what a good approximation of time for this would be.

@mattben What would you estimate, given that you wrote a script to pull from an old database before?

[durack1]

@James-Crean, just as a thought, the publication site/software would really benefit from the functionality that you're describing above, so provide a text file with comma-separated DOI entries for e.g., and the project that these entries will be uploaded into, and it just chugs along and does it automagically.. Would be a great feature to add..

[mauzey1]

@taylor13 @durack1

I have investigated reconstructing some of the functionality of the dynamic links. Using the old database's dump file, I made a JSON file containing the subproject details and publication abstracts for subproject_publications.html. I replaced the PHP files used to access the database with HTML+Javascript that parse the JSON file.

Although there are 603 "Pub" links on subproject_publications.html, there are only 385 entries in the publication database so almost half of those links will not show data. I'm not sure if the original database had more publications in it. Only 195 of the publication entries provide a DOI ID.

The dump file in the repo might have been made before more stuff was added. Is there a more recent version of the database dump file available?

I will make a pull request to add the new links to subproject_publications.html.

[mauzey1]

Here is what one of the subproject pages looks like.

A publication abstract page.

[durack1]

This is great, thanks for pushing on this @mauzey1!