From the graph's OP:

>Also for the language categorization, only the TITLE of the job offer has been analyzed. This means that for example, a title of "Backend developer" would be discarded, since it does not contain any language or stack valid on it. Analyzing only the title also filters out offers that require many languages and are fuzzy.

I get the negative pickiness about this sub, but he's really not being picky. The data is bad, suboptimal, and misleading. Most companies do not list job postings with very specific "language x developer" and sometimes put languages in the job details, so OP's very limited sampling method is a flaw. The last thing we want is a bunch of new CS students hyper focused on learning Solidify for the big money, but only to learn that most companies don't even know about that language or even allow answering technical interview questions with Solidfy.