c9fc7882e6
Seen on `github/codeql`, some queries had very poor performance:
```
[2/24 eval 36m4s] Evaluation done; writing results to
codeql/actions-queries/Security/CWE-312/ExcessiveSecretsExposure.bqrs
```
Investigating further lead to the following worrying sequence of joins
(after I ran out of patience and cancelled the query):
```
[2025-04-01 12:31:03] Tuple counts for
Yaml::YamlInclude.getTargetPath/0#dispred#32565107#fb#reorder_1_0/2@i6#9f4b2jw1
after 8m40s:
...
559418 ~33% {1} r5 = SCAN
`Yaml::YamlNode.getLocation/0#dispred#24555c57#prev_delta` OUTPUT In.1
...
909345525 ~821% {3} r7 = JOIN r5 WITH
`Yaml::YamlNode.getLocation/0#dispred#24555c57#prev` CARTESIAN PRODUCT
OUTPUT Rhs.1, Lhs.0 'result', Rhs.0
909342139 ~779% {3} | JOIN WITH
`Locations::Location.getFile/0#dispred#dcf38c8d#prev` ON FIRST 1 OUTPUT
Rhs.1, Lhs.1 'result', Lhs.2
909338753 ~794% {3} | JOIN WITH containerparent_10#join_rhs
ON FIRST 1 OUTPUT Rhs.1, Lhs.1 'result', Lhs.2
909335367 ~824% {3} | JOIN WITH
`FileSystem::Container.getAbsolutePath/0#dispred#d234e6fa` ON FIRST 1
OUTPUT Lhs.2, Lhs.1 'result', Rhs.1
883246724 ~812% {3} | JOIN WITH
`Yaml::YamlNode.getDocument/0#dispred#ee1eb3bf#bf_10#join_rhs` ON FIRST
1 OUTPUT Rhs.1 'this', Lhs.1 'result', Lhs.2
760047185 ~838% {5} | JOIN WITH yaml_scalars ON FIRST 1
OUTPUT Lhs.1 'result', Lhs.0 'this', Rhs.2, _, Lhs.2
0 ~0% {4} | REWRITE WITH Tmp.3 := "/", Out.3 :=
(In.4 ++ Tmp.3 ++ InOut.2), TEST Out.3 = InOut.0 KEEPING 4
{4} | REWRITE WITH NOT [TEST InOut.2
startsWith "/"]
...
```
The culprit turned out to be the following method on class `YamlInclude`
```ql
private string getTargetPath() {
exists(string path | path = this.getValue() |
if path.matches("/%")
then result = path
else
result =
this.getDocument().getLocation().getFile().getParentContainer().getAbsolutePath()
+ "/" +
path
)
}
```
Basically, in the `else` branch, the evaluator was producing all
possible values of `result` before filtering out the ones where the
`path` component started with a forward slash.
To fix this, I opted to factor out the logic into two helper predicates,
each accounting for whether `this.getValue()` does or does not start
with a `/`. With this, evaluating the original query from a clean cache
takes roughly 3.3s.
CodeQL
This open source repository contains the standard CodeQL libraries and queries that power GitHub Advanced Security and the other application security products that GitHub makes available to its customers worldwide.
How do I learn CodeQL and run queries?
There is extensive documentation about the CodeQL language, writing CodeQL using the CodeQL extension for Visual Studio Code and using the CodeQL CLI.
Contributing
We welcome contributions to our standard library and standard checks. Do you have an idea for a new check, or how to improve an existing query? Then please go ahead and open a pull request! Before you do, though, please take the time to read our contributing guidelines. You can also consult our style guides to learn how to format your code for consistency and clarity, how to write query metadata, and how to write query help documentation for your query.
For information on contributing to CodeQL documentation, see the "contributing guide" for docs.
License
The code in this repository is licensed under the MIT License by GitHub.
The CodeQL CLI (including the CodeQL engine) is hosted in a different repository and is licensed separately. If you'd like to use the CodeQL CLI to analyze closed-source code, you will need a separate commercial license; please contact us for further help.
Visual Studio Code integration
If you use Visual Studio Code to work in this repository, there are a few integration features to make development easier.
CodeQL for Visual Studio Code
You can install the CodeQL for Visual Studio Code extension to get syntax highlighting, IntelliSense, and code navigation for the QL language, as well as unit test support for testing CodeQL libraries and queries.
Tasks
The .vscode/tasks.json file defines custom tasks specific to working in this repository. To invoke one of these tasks, select the Terminal | Run Task... menu option, and then select the desired task from the dropdown. You can also invoke the Tasks: Run Task command from the command palette.