The front-end project serializes a representation of the CMS data using a GraphQL query. It is therefore important for the CMS team to be able to evaluate whether a given change that they've made has impacted the consistency of the CMS' response to that query.
The data is serialized as JSON in a file named pages.json
. When the command composer va:web:build
is executed, this file will be deposited in the folder (relative from the repository root) web/.cache/vagovdev/drupal/
.
So, by running the build command before and after making a change, we can compare the effect (if any) of the change on the data presented to the front-end by the CMS.
Lists of nodes as returned by the query are not always ordered; consequently even back-to-back queries of an unchanged database can result in wildly different data. It is therefore necessary to process the pages.json
file to make it more deterministic and yield results that can be usefully compared. jq
is an excellent tool for doing so.
The general workflow is just to create the diff (using e.g. diff pages-before.json pages-after.json
) and manually review the first few lines of differences, using one copy of pages.json
to review the context of each difference. Ideally, it should quickly become apparent that many of the differences are the same objects presented in a different order. By using a jq
expression, we can sort those objects consistently to create a more uniform result.
For instance, if we have a structure like this (greatly simplified):
{ "data": {
"alerts": {
"entities": [
{
"id": "324",
"name": "Foo"
},
{
"id": "523423",
"name": "Bar"
}
]
}
}
}
and the entities
objects appear in a different order, then we can use the following command to re-sort those objects by their "id" property:
# -S sorts the keys of objects upon output, which makes object-level diffs deterministic
#
# '.data.alerts.entities|=sort_by(.id)' sorts the objects within `data.alerts.entities`
# by their `id` property
jq -S '.data.alerts.entities|=sort_by(.id)'
And then these conflicts will disappear.
At the time of writing, the following commands were sufficient to create a "consistent" format with few differences from test to test:
jq -S '.data.alerts.entities|=sort_by(.id)' pages.json > pages2.json
jq -S '.data.nodeQuery.entities|=sort_by(.entityId)' pages2.json > pages3.json
After processing both the "before" and "after" pages.json
in this way, diffing the two should yield a far more useful visualization of the changes.