potential performance issue in the analyzer with large switch statements

[devoncarew]

I'm working on some generated code and am seeing a potential performance issue in the analyzer with large switch statements. I have a switch statement with 24,615 cases and notice that:

• the project takes about twice as long to analyze after I've generated this file than it did before
• when I open the file in vs code, the outline view takes a very long time to populate (even though when it finishes - 1-2 minutes later - the view only has ~200 symbols)

Perhaps there's some non-linear algorithm involving switch statements? The file itself is large - 1,379kb - but I'm working with other files that size or larger and not seeing the same issue.

[devoncarew]

To repro the issue, see the br_table.0.dart file here: https://github.com/devoncarew/wasmd/tree/main/test/spec/br_table.

[scheglov]

Yes, Observatory shows quite interesting picture.
There are so far two big performance issues visible:

1. FlowAnalysisHelper.getLabelTarget - probably indeed O(n^2)
2. FlowModel._clone - at the end there are 24618 in variableInfo map. It looks that each case in the switch adds one new entry to variableInfo, so it is also O(n^2).

[srawlins]

CC @stereotype441 for FlowModel performance.