I have experience / responsibility for creating dozens of language interfaces .
Talkative and punctuation languages are generally equally difficult for analysis and static analysis.
People who define languages of any kind, either decorate them for decades (for example, COBOL since 1958), or build complex languages (C ++, Scala, Ruby) with complex syntax and complex name resolution and rule type inference; compiler vendors then proceed to add obscure syntax to support the weird things they do, or to provide client locks (for example, "managed C ++" MS, DLL declarations, etc.). There is a third problem with lousy definitions; upper languages can have clear rules about how they work, but many languages have sloppy definitions (like PHP) that create dark angular cases that need to be eliminated due to painful experiments with the actual implementation.
C ++ was our worst, especially. with the C ++ 11 committee creating a massive recent mess of things. We have complete C ++ parsers, but they are still working on full name resolution for C ++ 11 on top of our implementation in C ++ 98. (The name resolution code is about 250,000 lines of code and not enough!).
IBM COBOL takes second place; the language is simply gigantic, and there are all kinds of ridiculous rules for resolving names ("an unqualified name can refer to a specific name without qualification if the link is unambiguous." So, is this name an unambiguous reference in this context?).
After you finish parsing and resolving the name / type, you get into the control flow, data flow, point analysis, anlaysis range, plotting calls ... which, as a rule, have the same amount of effort, earlier stages; we avoid less by having really good libraries that support these tasks.
With all of this as a background analysis, you can start doing a “static analysis" of the smart look that people want.
Another poster noted that recovery from syntax errors and (highlighting) "continue to generate meaningful error messages." All I can say is "Amen, brother." See this SO answer fooobar.com/questions/916357 / ... for a discussion of what goes wrong when you have "partial programs", which is essentially what you get when a syntax error correction guesses the correction .