[polskieradio] Add thumbnails.

[jakubadamw]

What is the purpose of your pull request?

• Bug fix
• New extractor
• New feature

---

Description of your pull request and other information

This is a simple change to extract thumbnails from Polskie Radio auditions.

[yan12125]

Usually the thumbnail field in _TESTS uses regular expressions instead of full URLs. See https://github.com/rg3/youtube-dl#adding-support-for-a-new-site for an example.

[jakubadamw]

@yan12125 Well, it's important that we check it finds the right image. Especially on Polskie Radio the page often contains, aside from the main image, multiple thumbnails for related materials.

[yan12125]

To avoid any possible false-positive, the extraction method itself should be precise. In this case, it's unlikely that _og_search_thumbnail finds a wrong image as in normal cases there's only one og:image tag in a web page.

[jakubadamw]

Agree, yet the test should not be concerned with the implementation details.

[yan12125]

You want to make sure that the extractor always gives the correct thumbnail, but using a full URL does not increase the detection rate of website changes. Thumbnail URLs are constantly changing. If there's a failed test, we can't know, without further investigation, whether it's due to wrong files or just moved/renamed files. Though I don't have a concrete statistical result, moved files are more common than wrong files, and in this case, a moved/renamed file is almost the only possible reason. As a result, if there's a failed test in polskieradio in the future, we will just update the thumbnail URL. It wastes time of developers, so I recommend you to use a regular expression.

If you really want a strict check, the correct way is to check the MD5 checksum of thumbnails, which is much less unlikely to change.

[jakubadamw]

That's a fair point on the MD5, thank you. I made the suggested change. :)

[yan12125]

Well md5: checks for the field value but not the corresponding file contents. You'll have to extend test/test_download.py

[jakubadamw]

@yan12125 That seems like a good idea, sadly I'm rather time constrained now. I've gone with the regular expression approach for now.

Is this good to go now?

[yan12125]

Thanks. Merged with minor fixes.

[jakubadamw]

Ah, of course. Ought to have escaped the dots. Thanks!