Error on implementation of SVG-Images from Wikimedia Commons

[LibrErli]

E.g. ZHB Luzern retrieves as example image the logo image stored as *svg on Wikimedia Commons. It seems that the lobid-imageproxy has troubles with implementing svg correctly.

• Image-Proxy-URL: https://lobid.org/imagesproxy?url=https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg?width=270
• FilePath: https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg

[dr0i]

The image seems to be of type PNG:

$ curl -L 'https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg?width=270' -o image
$ file image
image: PNG image data, 270 x 60, 8-bit colormap, non-interlaced

But lobid's imagesproxy tries to parse it, based on the file's name, as SVG - which consequently fails.

What we can do: change the imagesproxy to:
a) determine type of image by taking the content-type header into account (and not just by parsing the URL to determine get the file's suffix)
b) get rid of all HTML parameter (?.*) - that would leave us with curl -L 'https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg which is a SVG

As a hip-shot I would say b) is easier to do, but then we might not fix all scenarios. On the other hand, a) may also be problematic, if not all images have a content-type header. (b) could also be just a fallback if a URLs path is not ending with a dot + three characters (aka file extension).

[acka47]

What is the status of this, @Phu2 @dr0i ?

[Phu2]

+1 for b)

@dr0i Where does the parameter come from?

Sorry for the delay. Somehow i missed this issue.

[Phu2]

SVGs don't have a size. Requesting SVGs using width parameter make Wikimedia deliver PNGs.

[dr0i]

Parameter comes from depiction.thumbnail in https://lobid.org/gnd/4647881-4.json .
Looking at the data, it seems that if we would use depiction.id we would get the URL ending with the file extension. If that's always the case we could just switch to use this field (images would be a of greater size, but then the internet is fast enough and images are only served when using the web UI (few hits resulting from human interactions) and not the web API (possible millions of hits resulting from machine interactions)). WDYT @Phu2 ?

[Phu2]

Otherwise, it just needs this minor adjustment hbz/lobid@a5d0b23 in order to get the correct content type? But maybe i'm on the wrong track.

[dr0i]

@Phu2 I like your proposal also. Please merge and deploy 👍

[Phu2]

Deployed to stage.
Works now: https://stage.lobid.org/imagesproxy?url=https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg?width=270

Does not work yet: https://lobid.org/imagesproxy?url=https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg?width=270

@LibrErli @dr0i Please confirm. I'm happy to deploy to production afterwards. And sorry again for the delay!

[dr0i]

+1

[Phu2]

Deployed to production.
Now this also works: https://lobid.org/imagesproxy?url=https://commons.wikimedia.org/wiki/Special:FilePath/ZHB%20Luzern%20Logo%20farbig.svg?width=270

Fixed by hbz/lobid@e1083c9 (sorry, used wrong reference in commit messgae)