About SARIF support
SARIF (Static Analysis Results Interchange Format) is an OASIS Standard that defines an output file format. The SARIF standard is used to streamline how static analysis tools share their results. Code scanning supports a subset of the SARIF 2.1.0 JSON schema.
To upload a SARIF file from a third-party static code analysis engine, you'll need to ensure that uploaded files use the SARIF 2.1.0 version. GitHub will parse the SARIF file and show alerts using the results in your repository as a part of the code scanning experience. For more information, see "Uploading a SARIF file to GitHub." For more information about the SARIF 2.1.0 JSON schema, see sarif-schema-2.1.0.json.
If you're using GitHub Actions with the CodeQL analysis workflow, using the CodeQL runner, or using the CodeQL CLI, then the code scanning results will automatically use the supported subset of SARIF 2.1.0. For more information, see "Configuring code scanning for a repository", "Running CodeQL runner in your CI system", or "Installing CodeQL CLI in your CI system."
You can upload multiple SARIF files for the same commit, and display the data from each file as code scanning results. When you upload multiple SARIF files for a commit, you must indicate a "category" for each analysis. The way to specify a category varies according to the analysis method:
- Using the CodeQL CLI directly, pass the
--sarif-categoryargument to thecodeql database analyzecommand when you generate SARIF files. For more information, see "Configuring CodeQL CLI in your CI system." - Using GitHub Actions with
codeql-action/analyze, the category is set automatically from the workflow name and any matrix variables (typically,language). You can override this by specifying acategoryinput for the action, which is useful when you analyze different sections of a mono-repository in a single workflow. - Using GitHub Actions to upload results from other static analysis tools, then you must specify a
categoryinput if you upload more than one file of results for the same tool in one workflow. For more information, see "Uploading a code scanning analysis with GitHub Actions." - If you are not using either of these approaches, you must specify a unique
runAutomationDetails.idin each SARIF file to upload. For more information about this property, seerunAutomationDetailsobject below.
If you upload a second SARIF file for a commit with the same category and from the same tool, the earlier results are overwritten. However, if you try to upload multiple SARIF files for the same tool and category in a single GitHub Actions workflow run, the misconfiguration is detected and the run will fail.
GitHub uses properties in the SARIF file to display alerts. For example, the shortDescription and fullDescription appear at the top of a code scanning alert. The location allows GitHub to show annotations in your code file. For more information, see "Managing code scanning alerts for your repository."
If you're new to SARIF and want to learn more, see Microsoft's SARIF tutorials repository.
Providing data to track code scanning alerts across runs
Each time the results of a new code scan are uploaded, the results are processed and alerts are added to the repository. To prevent duplicate alerts for the same problem, code scanning uses fingerprints to match results across various runs so they only appear once in the latest run for the selected branch. This makes it possible to match alerts to the correct line of code when files are edited. The ruleID for a result has to be the same across analysis.
Reporting consistent filepaths
The filepath has to be consistent across the runs to enable a computation of a stable fingerprint. If the filepaths differ for the same result, each time there is a new analysis a new alert will be created, and the old one will be closed. This will cause having multiple alerts for the same result.
Including data for fingerprint generation
GitHub uses the partialFingerprints property in the OASIS standard to detect when two results are logically identical. For more information, see the "partialFingerprints property" entry in the OASIS documentation.
SARIF files created by the CodeQL analysis workflow, using the CodeQL runner, or using the CodeQL CLI include fingerprint data. If you upload a SARIF file using the upload-sarif action and this data is missing, GitHub attempts to populate the partialFingerprints field from the source files. For more information about uploading results, see "Uploading a SARIF file to GitHub."
If you upload a SARIF file without fingerprint data using the /code-scanning/sarifs API endpoint, the code scanning alerts will be processed and displayed, but users may see duplicate alerts. To avoid seeing duplicate alerts, you should calculate fingerprint data and populate the partialFingerprints property before you upload the SARIF file. You may find the script that the upload-sarif action uses a helpful starting point: https://github.com/github/codeql-action/blob/main/src/fingerprints.ts. For more information about the API, see "Upload an analysis as SARIF data."
Understanding rules and results
SARIF files support both rules and results. The information stored in these elements is similar but serves different purposes.
-
Rules are an array of
reportingDescriptorobjects that are included in thetoolComponentobject. This is where you store details of the rules that are run during analysis. Information in these objects should change infrequently, typically when you update the tool. -
Results are stored as a series of
resultobjects underresultsin therunobject. Eachresultobject contains details for one alert in the codebase. Within theresultsobject, you can reference the rule that detected the alert.
When you compare SARIF files generated by analyzing different codebases with the same tool and rules, you should see differences in the results of the analyses but not in the rules.
Specifying the root for source files
Code scanning interprets results that are reported with relative paths as relative to the root of the repository analyzed. If a result contains an absolute URI, the URI is converted to a relative URI. The relative URI can then be matched against a file committed to the repository.
You can provide the source root for conversion from absolute to relative URIs in one of the following ways.
checkout_pathinput to thegithub/codeql-action/analyzeactioncheckout_uriparameter to the SARIF upload API endpoint. For more information, see "Code scanning" in the REST API documentationinvocation.workingDirectory.uriproperty in the SARIF file
If you provide a source root, any location of an artifact specified using an absolute URI must use the same URI scheme. If there is a mismatch between the URI scheme for the source root and one or more of the absolute URIs, the upload is rejected.
For example, a SARIF file is uploaded using a source root of file:///github/workspace.
# Conversion of absolute URIs to relative URIs for location artifacts
file:///github/workspace/src/main.go -> src/main.go
file:///tmp/go-build/tmp.go -> file:///tmp/go-build/tmp.go
The file is successfully uploaded as both absolute URIs use the same URI scheme as the source root.
Validating your SARIF file
You can check a SARIF file is compatible with code scanning by testing it against the GitHub ingestion rules. For more information, visit the Microsoft SARIF validator.
Notes:
-
SARIF upload supports a maximum of 5000 results per upload. Any results over this limit are ignored. If a tool generates too many results, you should update the configuration to focus on results for the most important rules or queries.
-
For each upload, SARIF upload supports a maximum size of 10 MB for the
gzip-compressed SARIF file. Any uploads over this limit will be rejected. If your SARIF file is too large because it contains too many results, you should update the configuration to focus on results for the most important rules or queries.
Supported SARIF output file properties
If you use a code analysis engine other than CodeQL, you can review the supported SARIF properties to optimize how your analysis results will appear on GitHub.
Note: You must supply an explicit value for any property marked as "required". The empty string is not supported for required properties.
Any valid SARIF 2.1.0 output file can be uploaded, however, code scanning will only use the following supported properties.
sarifLog object
| Name | Description |
|---|---|
$schema | Required. The URI of the SARIF JSON schema for version 2.1.0. For example, https://json.schemastore.org/sarif-2.1.0.json. |
version | Required. Code scanning only supports SARIF version 2.1.0. |
runs[] | Required. A SARIF file contains an array of one or more runs. Each run represents a single run of an analysis tool. For more information about a run, see the run object. |
run object
Code scanning uses the run object to filter results by tool and provide information about the source of a result. The run object contains the tool.driver tool component object, which contains information about the tool that generated the results. Each run can only have results for one analysis tool.
| Name | Description |
|---|---|
tool.driver | Required. A toolComponent object that describes the analysis tool. For more information, see the toolComponent object. |
tool.extensions[] | Optional. An array of toolComponent objects that represent any plugins or extensions used by the tool during analysis. For more information, see the toolComponent object. |
invocation.workingDirectory.uri | Optional. This field is used only when checkout_uri (SARIF upload API only) or checkout_path (GitHub Actions only) are not provided. The value is used to convert absolute URIs used in physicalLocation objects to relative URIs. For more information, see "Specifying the root for source files." |
results[] | Required. The results of the analysis tool. Code scanning displays the results on GitHub. For more information, see the result object. |
toolComponent object
| Name | Description |
|---|---|
name | Required. The name of the analysis tool. Code scanning displays the name on GitHub to allow you to filter results by tool. |
version | Optional. The version of the analysis tool. Code scanning uses the version number to track when results may have changed due to a tool version change rather than a change in the code being analyzed. If the SARIF file includes the semanticVersion field, version is not used by code scanning. |
semanticVersion | Optional. The version of the analysis tool, specified by the Semantic Versioning 2.0 format. Code scanning uses the version number to track when results may have changed due to a tool version change rather than a change in the code being analyzed. If the SARIF file includes the semanticVersion field, version is not used by code scanning. For more information, see "Semantic Versioning 2.0.0" in the Semantic Versioning documentation. |
rules[] | Required. An array of reportingDescriptor objects that represent rules. The analysis tool uses rules to find problems in the code being analyzed. For more information, see the reportingDescriptor object. |
reportingDescriptor object
This is where you store details of the rules that are run during analysis. Information in these objects should change infrequently, typically when you update the tool. For more information, see "Understanding rules and results" above.
| Name | Description |
|---|---|
id | Required. A unique identifier for the rule. The id is referenced from other parts of the SARIF file and may be used by code scanning to display URLs on GitHub. |
name | Optional. The name of the rule. Code scanning displays the name to allow results to be filtered by rule on GitHub. |
shortDescription.text | Required. A concise description of the rule. Code scanning displays the short description on GitHub next to the associated results. |
fullDescription.text | Required. A description of the rule. Code scanning displays the full description on GitHub next to the associated results. The max number of characters is limited to 1000. |
defaultConfiguration.level | Optional. Default severity level of the rule. Code scanning uses severity levels to help you understand how critical the result is for a given rule. This value can be overridden by the level attribute in the result object. For more information, see the result object. Default: warning. |
help.text | Required. Documentation for the rule using text format. Code scanning displays this help documentation next to the associated results. |
help.markdown | Recommended. Documentation for the rule using Markdown format. Code scanning displays this help documentation next to the associated results. When help.markdown is available, it is displayed instead of help.text. |
properties.tags[] | Optional. An array of strings. Code scanning uses tags to allow you to filter results on GitHub. For example, it is possible to filter to all results that have the tag security. |
properties.precision | Recommended. A string that indicates how often the results indicated by this rule are true. For example, if a rule has a known high false-positive rate, the precision should be low. Code scanning orders results by precision on GitHub so that the results with the highest level, and highest precision are shown first. Can be one of: very-high, high, medium, or low. |
properties.problem.severity | Recommended. A string that indicates the level of severity of any alerts generated by a non-security query. This, with the properties.precision property, determines whether the results are displayed by default on GitHub so that the results with the highest problem.severity, and highest precision are shown first. Can be one of: error, warning, or recommendation. |
properties.security-severity | Recommended. A string representing a score that indicates the level of severity, between 0.0 and 10.0, for security queries (@tags includes security). This, with the properties.precision property, determines whether the results are displayed by default on GitHub so that the results with the highest security-severity, and highest precision are shown first. Code scanning translates numerical scores as follows: over 9.0 is critical, 7.0 to 8.9 is high, 4.0 to 6.9 is medium and 3.9 or less is low. |
result object
Each result object contains details for one alert in the codebase. Within the results object, you can reference the rule that detected the alert. For more information, see "Understanding rules and results" above.
Notes:
-
SARIF upload supports a maximum of 5000 results per upload. Any results over this limit are ignored. If a tool generates too many results, you should update the configuration to focus on results for the most important rules or queries.
-
For each upload, SARIF upload supports a maximum size of 10 MB for the
gzip-compressed SARIF file. Any uploads over this limit will be rejected. If your SARIF file is too large because it contains too many results, you should update the configuration to focus on results for the most important rules or queries.
| Name | Description |
|---|---|
ruleId | Optional. The unique identifier of the rule (reportingDescriptor.id). For more information, see the reportingDescriptor object. Code scanning uses the rule identifier to filter results by rule on GitHub. |
ruleIndex | Optional. The index of the associated rule (reportingDescriptor object) in the tool component rules array. For more information, see the run object. The allowed range for this property 0 to 2^63 - 1. |
rule | Optional. A reference used to locate the rule (reporting descriptor) for this result. For more information, see the reportingDescriptor object. |
level | Optional. The severity of the result. This level overrides the default severity defined by the rule. Code scanning uses the level to filter results by severity on GitHub. |
message.text | Required. A message that describes the result. Code scanning displays the message text as the title of the result. Only the first sentence of the message will be displayed when visible space is limited. |
locations[] | Required. The set of locations where the result was detected up to a maximum of 10. Only one location should be included unless the problem can only be corrected by making a change at every specified location. Note: At least one location is required for code scanning to display a result. Code scanning will use this property to decide which file to annotate with the result. Only the first value of this array is used. All other values are ignored. |
partialFingerprints | Required. A set of strings used to track the unique identity of the result. Code scanning uses partialFingerprints to accurately identify which results are the same across commits and branches. Code scanning will attempt to use partialFingerprints if they exist. If you are uploading third-party SARIF files with the upload-action, the action will create partialFingerprints for you when they are not included in the SARIF file. For more information, see "Providing data to track code scanning alerts across runs." Note: Code scanning only uses the primaryLocationLineHash. |
codeFlows[].threadFlows[].locations[] | Optional. An array of location objects for a threadFlow object, which describes the progress of a program through a thread of execution. A codeFlow object describes a pattern of code execution used to detect a result. If code flows are provided, code scanning will expand code flows on GitHub for the relevant result. For more information, see the location object. |
relatedLocations[] | A set of locations relevant to this result. Code scanning will link to related locations when they are embedded in the result message. For more information, see the location object. |
location object
A location within a programming artifact, such as a file in the repository or a file that was generated during a build.
| Name | Description |
|---|---|
location.id | Optional. A unique identifier that distinguishes this location from all other locations within a single result object. The allowed range for this property 0 to 2^63 - 1. |
location.physicalLocation | Required. Identifies the artifact and region. For more information, see the physicalLocation. |
location.message.text | Optional. A message relevant to the location. |
physicalLocation object
| Name | Description |
|---|---|
artifactLocation.uri | Required. A URI indicating the location of an artifact, usually a file either in the repository or generated during a build. For the best results we recommend that this is a relative path from the root of the GitHub repository being analyzed. For example, src/main.js. For more information about artifact URIs, see "Specifying the root for source files." |
region.startLine | Required. The line number of the first character in the region. |
region.startColumn | Required. The column number of the first character in the region. |
region.endLine | Required. The line number of the last character in the region. |
region.endColumn | Required. The column number of the character following the end of the region. |
runAutomationDetails object
The runAutomationDetails object contains information that specifies the identity of a run.
Note: runAutomationDetails is a SARIF v2.1.0 object. If you're using the CodeQL CLI, you can specify the version of SARIF to use. The equivalent object to runAutomationDetails is <run>.automationId for SARIF v1 and <run>.automationLogicalId for SARIF v2.
| Name | Description |
|---|---|
id | Optional. A string that identifies the category of the analysis and the run ID. Use if you want to upload multiple SARIF files for the same tool and commit, but performed on different languages or different parts of the code. |
The use of the runAutomationDetails object is optional.
The id field can include an analysis category and a run ID. We don't use the run ID part of the id field, but we store it.
Use the category to distinguish between multiple analyses for the same tool or commit, but performed on different languages or different parts of the code. Use the run ID to identify the specific run of the analysis, such as the date the analysis was run.
id is interpreted as category/run-id. If the id contains no forward slash (/), then the entire string is the run_id and the category is empty. Otherwise, category is everything in the string until the last forward slash, and run_id is everything after.
id | category | run_id |
|---|---|---|
| my-analysis/tool1/2021-02-01 | my-analysis/tool1 | 2021-02-01 |
| my-analysis/tool1/ | my-analysis/tool1 | no run-id |
| my-analysis for tool1 | no category | my-analysis for tool1 |
- The run with an
idof "my-analysis/tool1/2021-02-01" belongs to the category "my-analysis/tool1". Presumably, this is the run from February 2, 2021. - The run with an
idof "my-analysis/tool1/" belongs to the category "my-analysis/tool1" but is not distinguished from other runs in that category. - The run whose
idis "my-analysis for tool1 " has a unique identifier but cannot be inferred to belong to any category.
For more information about the runAutomationDetails object and the id field, see runAutomationDetails object in the OASIS documentation.
Note that the rest of the supported fields are ignored.
SARIF output file examples
These example SARIF output files show supported properties and example values.
Example with minimum required properties
This SARIF output file has example values to show the minimum required properties for code scanning results to work as expected. If you remove any properties, omit values, or use an empty string, this data will not be displayed correctly or sync on GitHub.
{
"$schema": "https://json.schemastore.org/sarif-2.1.0.json",
"version": "2.1.0",
"runs": [
{
"tool": {
"driver": {
"name": "Tool Name",
"rules": [
{
"id": "R01"
...
"properties" : {
"id" : "java/unsafe-deserialization",
"kind" : "path-problem",
"name" : "...",
"problem.severity" : "error",
"security-severity" : "9.8",
}
}
]
}
},
"results": [
{
"ruleId": "R01",
"message": {
"text": "Result text. This result does not have a rule associated."
},
"locations": [
{
"physicalLocation": {
"artifactLocation": {
"uri": "fileURI"
},
"region": {
"startLine": 2,
"startColumn": 7,
"endColumn": 10
}
}
}
],
"partialFingerprints": {
"primaryLocationLineHash": "39fa2ee980eb94b0:1"
}
}
]
}
]
}
Example showing all supported SARIF properties
This SARIF output file has example values to show all supported SARIF properties for code scanning.
{
"$schema": "https://json.schemastore.org/sarif-2.1.0.json",
"version": "2.1.0",
"runs": [
{
"tool": {
"driver": {
"name": "Tool Name",
"semanticVersion": "2.0.0",
"rules": [
{
"id": "3f292041e51d22005ce48f39df3585d44ce1b0ad",
"name": "js/unused-local-variable",
"shortDescription": {
"text": "Unused variable, import, function or class"
},
"fullDescription": {
"text": "Unused variables, imports, functions or classes may be a symptom of a bug and should be examined carefully."
},
"defaultConfiguration": {
"level": "note"
},
"properties": {
"tags": [
"maintainability"
],
"precision": "very-high"
}
},
{
"id": "d5b664aefd5ca4b21b52fdc1d744d7d6ab6886d0",
"name": "js/inconsistent-use-of-new",
"shortDescription": {
"text": "Inconsistent use of 'new'"
},
"fullDescription": {
"text": "If a function is intended to be a constructor, it should always be invoked with 'new'. Otherwise, it should always be invoked as a normal function, that is, without 'new'."
},
"properties": {
"tags": [
"reliability",
"correctness",
"language-features"
],
"precision": "very-high"
}
},
{
"id": "R01"
}
]
}
},
"automationDetails": {
"id": "my-category/"
},
"results": [
{
"ruleId": "3f292041e51d22005ce48f39df3585d44ce1b0ad",
"ruleIndex": 0,
"message": {
"text": "Unused variable foo."
},
"locations": [
{
"physicalLocation": {
"artifactLocation": {
"uri": "main.js",
"uriBaseId": "%SRCROOT%"
},
"region": {
"startLine": 2,
"startColumn": 7,
"endColumn": 10
}
}
}
],
"partialFingerprints": {
"primaryLocationLineHash": "39fa2ee980eb94b0:1",
"primaryLocationStartColumnFingerprint": "4"
}
},
{
"ruleId": "d5b664aefd5ca4b21b52fdc1d744d7d6ab6886d0",
"ruleIndex": 1,
"message": {
"text": "Function resolvingPromise is sometimes invoked as a constructor (for example [here](1)), and sometimes as a normal function (for example [here](2))."
},
"locations": [
{
"physicalLocation": {
"artifactLocation": {
"uri": "src/promises.js",
"uriBaseId": "%SRCROOT%"
},
"region": {
"startLine": 2
}
}
}
],
"partialFingerprints": {
"primaryLocationLineHash": "5061c3315a741b7d:1",
"primaryLocationStartColumnFingerprint": "7"
},
"relatedLocations": [
{
"id": 1,
"physicalLocation": {
"artifactLocation": {
"uri": "src/ParseObject.js",
"uriBaseId": "%SRCROOT%"
},
"region": {
"startLine": 2281,
"startColumn": 33,
"endColumn": 55
}
},
"message": {
"text": "here"
}
},
{
"id": 2,
"physicalLocation": {
"artifactLocation": {
"uri": "src/LiveQueryClient.js",
"uriBaseId": "%SRCROOT%"
},
"region": {
"startLine": 166
}
},
"message": {
"text": "here"
}
}
]
},
{
"ruleId": "R01",
"message": {
"text": "Specifying both [ruleIndex](1) and [ruleID](2) might lead to inconsistencies."
},
"level": "error",
"locations": [
{
"physicalLocation": {
"artifactLocation": {
"uri": "full.sarif",
"uriBaseId": "%SRCROOT%"
},
"region": {
"startLine": 54,
"startColumn": 10,
"endLine": 55,
"endColumn": 25
}
}
}
],
"relatedLocations": [
{
"id": 1,
"physicalLocation": {
"artifactLocation": {
"uri": "full.sarif"
},
"region": {
"startLine": 81,
"startColumn": 10,
"endColumn": 18
}
},
"message": {
"text": "here"
}
},
{
"id": 2,
"physicalLocation": {
"artifactLocation": {
"uri": "full.sarif"
},
"region": {
"startLine": 82,
"startColumn": 10,
"endColumn": 21
}
},
"message": {
"text": "here"
}
}
],
"codeFlows": [
{
"threadFlows": [
{
"locations": [
{
"location": {
"physicalLocation": {
"region": {
"startLine": 11,
"endLine": 29,
"startColumn": 10,
"endColumn": 18
},
"artifactLocation": {
"uriBaseId": "%SRCROOT%",
"uri": "full.sarif"
}
},
"message": {
"text": "Rule has index 0"
}
}
},
{
"location": {
"physicalLocation": {
"region": {
"endColumn": 47,
"startColumn": 12,
"startLine": 12
},
"artifactLocation": {
"uriBaseId": "%SRCROOT%",
"uri": "full.sarif"
}
}
}
}
]
}
]
}
],
"partialFingerprints": {
"primaryLocationLineHash": "ABC:2"
}
}
],
"columnKind": "utf16CodeUnits"
}
]
}