Welcome to mirror list, hosted at ThFree Co, Russian Federation.

gitlab.com/gitlab-org/gitlab-foss.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat (limited to 'doc/ci/yaml/includes.md')
-rw-r--r--doc/ci/yaml/includes.md83
1 files changed, 82 insertions, 1 deletions
diff --git a/doc/ci/yaml/includes.md b/doc/ci/yaml/includes.md
index b1f43a4679b..bf0b7444e78 100644
--- a/doc/ci/yaml/includes.md
+++ b/doc/ci/yaml/includes.md
@@ -157,6 +157,87 @@ and the `environment:url` of the `production` job defined in the `.gitlab-ci.yml
override the values defined in the `autodevops-template.yml` file. The other keywords
do not change. This method is called *merging*.
+### Merge method for `include`
+
+For a file containing `include` directives, the included files are read in order (possibly
+recursively), and the configuration in these files is likewise merged in order. If the parameters overlap, the last included file takes precedence. Finally, the directives in the
+file itself are merged with the configuration from the included files.
+
+This merge method is a _deep merge_, where hash maps are merged at any depth in the
+configuration. To merge hash map A (containing the configuration merged so far) and B (the next piece
+of configuration), the keys and values are processed as follows:
+
+- When the key only exists in A, use the key and value from A.
+- When the key exists in both A and B, and their values are both hash maps, merge those hash maps.
+- When the key exists in both A and B, and one of the values is not a hash map, use the value from B.
+- Otherwise, use the key and value from B.
+
+For example:
+
+We have a configuration consisting of two files.
+
+- The `.gitlab-ci.yml` file:
+
+ ```yaml
+ include: 'common.yml'
+
+ variables:
+ POSTGRES_USER: username
+
+ test:
+ rules:
+ - if: $CI_PIPELINE_SOURCE == "merge_request_event"
+ when: manual
+ artifacts:
+ reports:
+ junit: rspec.xml
+ ```
+
+- The `common.yml` file:
+
+ ```yaml
+ variables:
+ POSTGRES_USER: common_username
+ POSTGRES_PASSWORD: testing_password
+
+ test:
+ rules:
+ - when: never
+ script:
+ - echo LOGIN=${POSTGRES_USER} > deploy.env
+ - rake spec
+ artifacts:
+ reports:
+ dotenv: deploy.env
+ ```
+
+The merged result:
+
+```yaml
+variables:
+ POSTGRES_USER: username
+ POSTGRES_PASSWORD: testing_password
+
+test:
+ rules:
+ - if: $CI_PIPELINE_SOURCE == "merge_request_event"
+ when: manual
+ script:
+ - echo LOGIN=${POSTGRES_USER} > deploy.env
+ - rake spec
+ artifacts:
+ reports:
+ junit: rspec.xml
+ dotenv: deploy.env
+```
+
+In this example:
+
+- Variables are only evaluated after all the files are merged together. A job in an included file
+ might end up using a variable value defined in a different file.
+- `rules` is an array so it cannot be merged. The top-level file takes precedence.
+- `artifacts` is a hash map so it can be deep merged.
+
## Override included configuration arrays
You can use merging to extend and override configuration in an included template, but
@@ -304,7 +385,7 @@ In GitLab 14.5 and later, you can also use:
- [Trigger variables](../triggers/index.md#pass-cicd-variables-in-the-api-call).
- [Scheduled pipeline variables](../pipelines/schedules.md#add-a-pipeline-schedule).
-- [Manual pipeline run variables](../variables/index.md#override-a-variable-when-running-a-pipeline-manually).
+- [Manual pipeline run variables](../pipelines/index.md#run-a-pipeline-manually).
- Pipeline [predefined variables](../variables/predefined_variables.md).
YAML files are parsed before the pipeline is created, so the following pipeline predefined variables