Images in instant answer are not loaded if "image proxy" is enables

[catfluoride]

SearXNG Version: 1.0.0-1669-10e6881c

Simple theme is used

How did you install SearXNG?

SearxNG was installed by "installation scripts", without Morty or Filtron.

What happened?

If "image proxy" is enabled, images searches work correctly. However, images in "Instant Answers" under the "General" category are absent and just a broken image placeholder is shown.

For instance, if I search for "Enrico Fermi" in the General category, with the Image Proxy enabled Wikipedia's image is broken.

How To Reproduce
This bug is reproducible. Just enable "Image Proxy" an settings GUI and perform the aforementioned search

Expected behavior

An image should be displayed in place of the broken image placeholder

Screenshots & Logs

Expected result:

Actual result:

Inspection of the page source gives the following:

<img src="/image_proxy?url=https%3A%2F%2Fcommons.wikimedia.org%2Fwiki%2FSpecial%3AFilePath%2FEnrico%2520Fermi%25201943-49.jpg%3Fwidth%3D500%26height%3D400&amp;h=dff389296b41618681172e6ca9f6a4e50b6ad32554484ff828889e64f4a419ab" title="Enrico Fermi" alt="Enrico Fermi">

Somewhat "image_proxy" is misbehaving on the home page.

[return42]

I can confirm that the image is not loaded:

INFO    werkzeug                      : 127.0.0.1 - - [06/Feb/2022 12:24:41] "GET /static/themes/simple/js/searxng.min.js HTTP/1.1" 304 -
DEBUG   searx.network.image_proxy     : HTTP Request: GET https://commons.wikimedia.org/wiki/Special:FilePath/Enrico%20Fermi%201943-49.jpg?width=500&height=400 "HTTP/1.1 302 Found" (text/html; charset=utf-8)
DEBUG   searx.webapp                  : image-proxy: wrong response code: 302
INFO    werkzeug                      : 127.0.0.1 - - [06/Feb/2022 12:24:41] "GET /image_proxy?url=https%3A%2F%2Fcommons.wikimedia.org%2Fwiki%2FSpecial%3AFilePath%2FEnrico%2520Fermi%25201943-49.jpg%3Fwidth%3D500%26height%3D400&h=c5affd8c6746b4d64eb63447da9f4138674fcab7dbc0a60576267cbea2232578 HTTP/1.1" 400 -

But the URL (now without image_proxy) ...

is valid: https://commons.wikimedia.org/wiki/Special:FilePath/Enrico%20Fermi%201943-49.jpg?width=500&height=400

[tiekoetter]

I can't reproduce this.

If wikidata is enabled the image is broken.

See https://searx.tiekoetter.com/search?q=%21wikidata+enrico+fermi&language=en-US&time_range=&safesearch=0&theme=simple

[catfluoride]

I can't reproduce this.

If wikidata is enabled the image is broken.

See https://searx.tiekoetter.com/search?q=%21wikidata+enrico+fermi&language=en-US&time_range=&safesearch=0&theme=simple

Indeed, this it it. Disabling wikidata, everything works as expected on my instance.
Thank you for sharing this finding.

[tiekoetter]

@return42 Does the image proxy have a size limit?

[return42]

It seems the img URL from Wikidata's PNG is redirected (twice)

https://commons.wikimedia.org/wiki/Special:FilePath/Enrico%20Fermi%201943-49.jpg?width=500&height=400

while the URL of the JPG from wikipedia is not redirected

https://upload.wikimedia.org/wikipedia/commons/thumb/d/d4/Enrico_Fermi_1943-49.jpg/258px-Enrico_Fermi_1943-49.jpg

[tiekoetter]

I am currently working on a solution. I think I can do a PR soon.

[return42]

I am currently working on a solution. I think I can do a PR soon.

Thanks .. I also working on a solution ;-) .. I just stuck at the meaning of priority . and why it must be lower:

searxng/searx/engines/wikidata.py

Lines 223 to 225 in 10e6881

	if attribute.priority < img_src_priority:
	img_src = value
	img_src_priority = attribute.priority

---

EDIT: I think image proxy should support redirects

[tiekoetter]

EDIT: I think image proxy should support redirects

Maybe but then image proxy would do 2 redirects for every wikidata image. I don't think this would be good for performance.

[return42]

Maybe but then image proxy would do 2 redirects for every wikidata image. I don't think this would be good for performance.

This arguments counts .. I just think about rewrite the image URL in the wikidata engine ... needs some more analysis .. I just digging :)

[return42]

OK, if I'm not wrong all images from commons.wikimedia.org are redirected to upload.wikimedia.org .. may be it is better we do not use images from commons.wikimedia.org in the info box.

Another issue seems the lower priority condition in

searxng/searx/engines/wikidata.py

Lines 223 to 225 in 10e6881

	if attribute.priority < img_src_priority:
	img_src = value
	img_src_priority = attribute.priority

[tiekoetter]

@return42
tiekoetter@4c8a6cd
This should temporarily fix wikidata until a better solution can be found.

[return42]

@tiekoetter: I cloned your branch .. wow, very cool ..

img_src_name_md5[0] + "/" + img_src_name_md5[0:2]

from where did you know it? I have never thought that it is the prefix of md5 digest.

Do you like to send a PR .. in the PR we can do some polish, but the md5 digest is the solution I think.

[tiekoetter]

from where did you know it? I have never thought that it is the prefix of md5 digest.

https://stackoverflow.com/questions/33689980/get-thumbnail-image-from-wikimedia-commons

I will send the PR now.