Wikimedia Commons: Image file doesn't attach

[adam3smith]

Example: https://commons.wikimedia.org/wiki/File:Baryon_decuplet.png Reported: https://forums.zotero.org/discussion/107766/unable-to-download-file-from-wikiedia-commons#latest

To be clear: the metadata imports fine, as does the snapshot, by the translator is supposed to also download the image on the page, which it doesn't anymore.

I haven't looked at this in any more detail, but wouldn't be surprising if Commons changed how image files are made available.

[AbeJellinek]

@zoe-translates: Apparently this translator still uses FW - any interest in rewriting?

[zoe-translates]

This is the most recently-updated translator that still relies on FW. I'm going to stop that :)

One question about downloading the images themselves -

Some files hosted by the Wikimedia Commons could be super large. Should we impose an upper-limit on the attachment's file size, above which the file is saved as a link (snapshot: false) rather than a download?

[AbeJellinek]

Should we impose an upper-limit on the attachment's file size, above which the file is saved as a link (snapshot: false) rather than a download?

Sure. 10 MB?

[zoe-translates]

In fact, the reported problem was caused by the xpath here (and the second appearance on line 121:

https://github.com/zotero/translators/blob/8e5c648bb1e2ec58eccbc5b04ab5d1e1e656afdc/Wikimedia%20Commons.js#L85-L88

The xpath //div[@class="fullMedia"]//a[@class="internal"] would have matched the a tag.

However there are a lot more problems with this translator that I'm unwilling to submit just this change of xpath as a hotfix (it won't actually "fix" anything, because the other metadata fields are still broken, and the attachment itself misses mimeType).

So I think this is an opportunity to re-write the translator.

[zoe-translates]

Todo:

• [ ] Better type determination (not just "artwork" for everything; other content types: book, audio recording, video, or even "document"?)
• [ ] Determine mimeType and file size (maybe complicated by interface language (e.g. "Mio" for mébioctet vs. "MB")
• [ ] What actually goes in the title? For example, using the current FW-based translator, the page https://commons.wikimedia.org/wiki/File:The_Day_the_Earth_Smiled_-_PIA17172.jpg produces the title "English: On July 19, 2013, in an event celebrated the world over, NASA's Cassini spacecraft slipped into Saturn's shadow and turned to image the planet, seven of its moons, its inner rings -- and, in the background, our home planet, Earth." None of these appears in the "cite this page" output provided by WMC itself
• [ ] Authorship: should I simply use "Wikimedia Commons contributors" as recommended by the "cite this page" output, or go for the actual author of an artwork (e.g. Caravaggio, for the first test case)? I think it's OK to cite "Wikimedia Commons contributors" if the item is treated as a web page (which happens to host artwork), but the original intent seems to be going for the artwork itself by piercing the hosting-on-web-page veil (see also this "How To Credit Images Found in the Wikimedia Commons")
• [ ] Are there more structured way to do this?