Purpose

While adding better validations for config options and per-snapshot validations, data must be validated against schemas other than the global config schema. To facilitate this, the validate util was refactored slightly. The merge util is also used within the core snapshot options config function to merge and validate provided options with config options and defaults while also handling deprecated options. Due to the merge util's limitation of only merging two objects at a time while mutating the first object, it had to be used several times in a row in order to merge all of the snapshot options together into a new object. Furthermore, it's options argument was limited in it's abilities and the merge function itself was centered around config normalization, so wasn't generic enough to use outside of said normalization function.

Approach

The validate utils were refactored so default schema $ids now start with a forward slash (/) to help distinguish them from local schema references. The function will also now only return an array of errors on failure and undefined on success. The associated addSchema util was updated to accept schemas without automatically adding them to the global Percy config schema (via an explicit $id). This allows the validate util to be used with other registered schemas and not just the global Percy config schema.

The merge util was refactored to accept an array of objects instead of only two objects. It also always returns a deep clone rather than mutating the first argument. The limited options argument was replaced with a map callback argument. The callback will be called for each property within each object, recursively, with the path, previous value, and next value of the property. A path-value pair can be returned to override the default merge behavior. Some options previously handled by the old merge function were replaced by the new map callback in the appropriate places.

Other features were also added while refactoring that were found to be useful for better validations:

• Schemas using minimum and maximum keywords will be clamped rather than scrubbed.
• Schemas can use a custom instanceof keyword to allow functions or regular expressions.
• Schemas can also use a custom disallowed keyword, based on the built-in required keyword (but negated).
• Nested objects that have missing required properties are now scrubbed.
• Empty properties such as array items that are missing or deleted are now scrubbed.
  e.g. ['foo', <5 empty items>, 'xyzzy'] => ['foo', 'xyzzy']
• Validation error paths were updated to show as proper array notation.
  before: snapshots.widths.0: must be <= 2000
  after: snapshots.widths[0]: must be <= 2000