Usage of print ast ql file

[Naman-ntc]

Print AST command is used in vscode-codeql extension where it uses the printast query in qlpack and displays the results.
I wanted to know how to get the ast outside of the vscode-extension?

printAst.ql implements a configuration to print ast for a single file (with selectedSourceFile) but it is unclear how to use this configuration

[hvitved]

This query is specifically written to be executed in the context of VS Code (because of the external string selectedSourceFile predicate). If you want to run it externally, you can make your own query:

import javascript
import semmle.javascript.PrintAst
import definitions

will print out the AST for all files, but you could also restrict it

import javascript
import semmle.javascript.PrintAst
import definitions

private File relevantSourceFile() { ... }

class PrintAstConfigurationOverride extends PrintAstConfiguration {
  override predicate shouldPrint(Locatable e, Location l) {
    super.shouldPrint(e, l) and
    l.getFile() = relevantSourceFile()
  }
}

[hvitved]

You can use getRelativePath or getAbsolutePath in combination with matches, see e.g. https://github.com/github/codeql/blob/main/csharp/ql/test/library-tests/controlflow/graph/Common.qll#L4.

[Naman-ntc]

On running the print-ast query and decoding the bqrs with cli - I am able to get a good ast in JSON. However, the nodes in this json only contain

• AST node type
• key (semmle.order or semmle.label)
• value (code string)

i.e. it does not provide the location of AST nodes in the javascript file. While I can write some program that matches the nodes from AST JSON file to code in the javascript, I believe some command/function should exist that gives line and column numbers for the nodes as the vs-code extension also links the nodes to the code files.

EDIT: Also it seems bqrs interpret with dot (graph-like output formats) is more informative than bqrs decode to a json. Still don't get line and column numbers though

[hvitved]

The following should work:

codeql query run -d <db> <query> -o output.bqrs
codeql bqrs decode --entities url --entities string  output.bqrs -o output.txt

[Naman-ntc]

Oh wow, that works like a charm. Thanks a lot for all the help!
Why isn't entities=all default in bqrs decode? This is somewhat confusing behavior

[hvitved]

Not sure why the default is what it is, but note that --entities all will include an additional id column, which is an internal (integer) representation of the entities, so this is mostly useful for internal testing.

[nakhlarafi]

Hello there,

Can I get the AST from the AST viewer in the VScode extension as a file?