example of an AST misalignment.

Rendered in GH

example of an token misalignment.

Rendered in GH

snapshot which provides a centralised list of fixtures with AST differences

snapshot which provides a centralised list of fixtures with Token differences

snapshot which provides a centralised list of fixtures with error differences (not error messages, just existence of errors)

This functionally behaves about the same asstr::replaceAll("'", '\\'), no?

I think you could technically write this as

str.replaceAll("\\","\\\\").replaceAll("'","\\'");

Though that would be slower because you'd need to do it in two passes over the string.

I haven't perf tested it, but given that there are going to be hundreds of these strings - each with a few hundred characters, I think that double handling will add up.

This feels like something we (@JamesHenry 😄) should feature request into Nx: the ability to have technical circular dependencies only in tests...

I can't remember the exact history of this but I'm pretty sure I provided this breakdown and created this solution to the problem.

As it turns out on this PR right now the@typescript-eslint/typescript-estree package is not in the package.json of ast-spec, so we could simplify things to:

The specifying of the package in the package.json was the real issue, and is the thing that causes the rigid circular dependency to exist between the nodes on the graph - there would be no way to know that the package was only used in certain situations when that is the way the relationship is defined. As part of the Nx updates I'm doing we'll be moving entirely to source to source relationships between packages/graph nodes (as is far more common) and we won't need to worry about this in that situation

(I didn't include it in the screenshot but feel free to move some of the commentary around why wedon't want to include the package in the package.json to make that lint rule happy and that's why we are disabling it inline)

Was there an update to nx to allow this? I swear there was a reason we went with using an.nxignore file.

Or did we used to rely on implicit dependency graph and now we have an explicit one configured it doesn't matter?

I'm afraid I don't 100% remember 😬

@bradzacher this seem to not be working on windows machines

Error: spawn yarn ENOENT    at Process.ChildProcess._handle.onexit (node:internal/child_process:283:19)    at onErrorNT (node:internal/child_process:478:16)    at processTicksAndRejections (node:internal/process/task_queues:83:21) {  errno: -4058,  code: 'ENOENT',  syscall: 'spawn yarn',  path: 'yarn',  spawnargs: [ 'build' ]}

and rolling this code back toconst execAsync = promisify(childProcess.exec); seem to be fixing it, not sure what is actually causing this issue

If you can figure out how to make it work on Windows and display the error logs - happy to accept a PR!

I'd assume the path isn't properly set so the binary can't be found? That's generally what ENONET means I think.

from my understunding of the issue, its related to/ and\ beeing incorrectly used, i did my tests on 2 env win 11 (stable) and win 11 eac (preview), with both node 16 and 18, i was able to reproduce this in all those cases
as for linux subsystem (ubuntu) (linux "vm" in windows) i was unable to reproduce this

edit:
shell, spawn option is required to correctly execute yarn instance (yarn.bat) on windows machines
see:https://stackoverflow.com/questions/37459717/error-spawn-enoent-on-windows

yarn isn't recognized in the console as a program but as a command.