5 ms·
Yes, you are right, I am a student who is starting to do static analysis, but it doesn't annoy me) I also want you to say that I am not analyzed here as part of
by l3r8y 4y ago
Yes, you are right, I am a student who is starting to do static analysis, but it doesn't annoy me) I also want you to say that I am not analyzed here as part of the general irony, I don't think anyone should take it seriously, anyway if you can write me what you think makes language difficult to analyze, I would love to read it and add it)
- WastingMyTime89 4y agoI am not sure I understood the second part of your sentence but sorry if what I wrote sounded harsh. It's easy to forget that you might be read by the person who actually wrote the article. I don't think that what you wrote is worthless. I just disagreed. > anyway if you can write me what you think makes language difficult to analyze, I would love to read it First, I can give you exemples of things which are notoriously hard to analyse in any language: order of execution dependent behaviours when parallelism is involved, floating point arithmetic, heavily runtime dependent behaviours. Regarding languages features, anything which limits the amount of guarantee you can give about semantic units make analysis more difficult (well, a lot less useful at least). Languages were non local changes can heavily influence what's happening are notoriously hard to analyse for exemple. As a worst case scenario, you can consider a language with no namespacing, global scoping, global visibility, global accessibility and first-class functions (it sounds silly but I have used languages for symbolic computation which work like that). Now, you have to consider the whole execution context to know the value of identifiers including functions. You can give no garantee whatsoever on the value of anything in isolation.
- l3r8y 4y agoThe number of comments with amendments forced me to do a second part. I will try to read and take into account all that have written today, thank you again)