CUE in Perses#
Perses comes with validation capabilities based on CUE, a powerful validation language that permitted us to move the type constraints out of Golang (static language), which in the end makes it possible to modify at runtime the list of objects that Perses can accept. More concretely, this allows to support new kinds of charts and/or datasources dynamically, in the form of plugins written in CUE.
Writing plugins#
This section explains about the format any plugin should follow to be accepted & registered by Perses at runtime.
Variable#
A variable plugin looks like the following:
package model
kind: "<Variable name>" // e.g kind: "PrometheusLabelValuesVariable"
spec: close({
labelName: string
matchers: [...string]
})
it should define:
- the
model
package, - the variable's
kind
, - the variable's
spec
containing any field you want for this variable plugin.
Panel#
A panel plugin looks like the following:
package model
kind: "<Panel name>" // e.g kind: "TimeSeriesChart",
spec: {
queries: [...#ts_query]
legend?: #legend
format?: common.#format
thresholds?: common.#thresholds
}
it should define:
- the
model
package, - the panel's
kind
, - the panel's
spec
containing any field you want for this panel plugin.
Query#
A query plugin looks like the following:
package model
kind: "<Query name>" // e.g kind: "PrometheusTimeSeriesQuery"
spec: {
datasource: {
kind: "<Datasource type>" // e.g kind: "PrometheusDatasource"
}
query: string
minStep?: =~"^(?:(\\d+)y)?(?:(\\d+)w)?(?:(\\d+)d)?(?:(\\d+)h)?(?:(\\d+)m)?(?:(\\d+)s)?(?:(\\d+)ms)?$"
resolution?: number
}
it should define:
- the
model
package, - the query's
kind
, - the query's
spec
containing:- a
datasource
field that holds thekind
of datasource corresponding to this query type, - any other field you want for this query plugin.
- a
Migration from Grafana#
A Perses plugin can optionally embed a migrate
folder file at its root, that contains a migrate.cue
file. This file is basically describing in CUE language how to convert a given Grafana object into an instance of this plugin. In such case your plugin is considered as the Perses equivalent of this Grafana object type, i.e it will be used as part of the translation process when a Grafana dashboard is received on the /api/migrate
endpoint.
Warning
If ever you come to the situation where you have 2 or more plugins describing a migration logic for the same Grafana panel type, be aware that the last one encountered by alphabetical order will take priority.
Variable#
A variable migration file looks like the following:
if #var.type == "custom" || #var.type == "interval" {
kind: "StaticListVariable"
spec: {
values: strings.Split(#var.query, ",")
}
},
- The file is named
migrate.cue
. - The file content is made of one or more conditional block(s), separated by commas (even if you have only one).
- Each conditional block defines one or more matches on attributes from the
#var
definition.#var
references a variable object from Grafana. You can access the different fields with like#var.field.subfield
. To know the list of fields available, check the Grafana datamodel for the considered variable type (from Grafana repo, or by inspecting the JSON of the dashboard on the Grafana UI).- You most certainly want a check on the
#var.type
value like shown in above example.
- Each conditional block contains a list of fields & assignments, meeting the requirements of the considered Perses variable plugin. Use the
#var.field.subfield
syntax to access the values from the Grafana variable, thus achieve its remapping into Perses.
Panel#
A panel migration file looks like the following:
package migrate
import (
commonMigrate "github.com/perses/perses/cue/schemas/common/migrate"
)
#grafanaType: "bargauge"
#panel: _
kind: "BarChart"
spec: {
calculation: *commonMigrate.#mapping.calc[#panel.options.reduceOptions.calcs[0]] | commonMigrate.#defaultCalc // only consider [0] here as Perses's GaugeChart doesn't support individual calcs
#unit: *commonMigrate.#mapping.unit[#panel.fieldConfig.defaults.unit] | null
if #unit != null {
format: unit: #unit
}
#decimal: *#panel.fieldConfig.defaults.decimal | null
if #decimal != null {
format: decimalPlaces: #decimal
}
}
- The file must be named
migrate.cue
. #grafanaType
is a mandatory definition to provide, whose string value must match thetype
of the Grafana panel you want to migrate.#panel
is the reference used by Perses to inject the Grafana panel objects to migrate.- You can access the different fields via the
#panel.field.subfield
syntax. To find the list of available fields, refer to the Grafana data model for the relevant panel type (from Grafana repo, or by inspecting the JSON of the dashboard on the Grafana UI). - Declaring
#panel: _
like in the above example is optional, it's just there to enable standalone validation of the file (_
means "any" in CUE).
- You can access the different fields via the
- The file consists of field assignments, using the content of
#panel
. The end result must match the model of the considered Perses panel plugin.- Optionally, you can use the
github.com/perses/perses/cue/schemas/common/migrate
package that Perses provides in order to remap some of the attributes:#mapping.unit
: mapping table for theunit
attribute (key = grafana unit, value = perses equivalent).#mapping.calc
: mapping table for thecalculation
attribute (key = grafana unit, value = perses equivalent).#mapping.color
: mapping table for the "standard" colors used by Grafana (key = color name, value = hex code).#defaultCalc
: standard default value for thecalculation
attribute.
- Optionally, you can use the
Query#
A query migration file looks like the following:
package migrate
#target: _
// NB we would need `if` to support short-circuit in order to avoid code duplication here.
// See https://github.com/cue-lang/cue/issues/2232
if (*#target.datasource.type | null) == "prometheus" && #target.expr != _|_ {
kind: "PrometheusTimeSeriesQuery"
spec: {
datasource: {
kind: "PrometheusDatasource"
name: #target.datasource.uid
}
query: #target.expr
#legendFormat: *#target.legendFormat | "__auto"
if #legendFormat != "__auto" {
seriesNameFormat: #legendFormat
}
if #target.interval != _|_ {
minStep: #target.interval
}
}
},
- The file must be named
migrate.cue
. #target
is the reference used by Perses to inject the Grafana target objects to migrate.- You can access the different fields via the
#target.field.subfield
syntax. To find the list of available fields, refer to the Grafana data model for the targets (from Grafana repo, or by inspecting the JSON of the dashboard on the Grafana UI). - Declaring
#target: _
like in the above example is optional, it's just there to enable standalone validation of the file (_
means "any" in CUE).
- You can access the different fields via the
- The migration logic must be wrapped into one or more conditional block(s). For each of them:
- The condition is about one or more attributes from the
#target
definition. You most certainly want a check on the#target.datasource.type
value like shown in above example. - The body consists of field assignments, using the content of
#target
. The end result must match the model of the considered Perses query plugin.
- The condition is about one or more attributes from the
Warning
Ensure that your file evaluates to an empty result if the provided #target
value does not match the expected type.
Variable#
A variable migration file looks like the following:
package migrate
import "strings"
#var: _
if #var.type == "custom" || #var.type == "interval" {
kind: "StaticListVariable"
spec: {
values: strings.Split(#var.query, ",")
}
}
- The file must be named
migrate.cue
. #var
is the reference used by Perses to inject the Grafana variable objects to migrate.- You can access the different fields via the
#var.field.subfield
syntax. To find the list of available fields, refer to the Grafana data model for the relevant variable type (from Grafana repo, or by inspecting the JSON of the dashboard on the Grafana UI). - Declaring
#var: _
like in the above example is optional, it's just there to enable standalone validation of the file (_
means "any" in CUE).
- You can access the different fields via the
- The migration logic must be wrapped into one or more conditional block(s). For each of them:
- The condition is about one or more attributes from the
#var
definition. You most certainly want a check on the#var.type
value like shown in above example. - The body consists of field assignments, using the content of
#var
. The end result must match the model of the considered Perses variable plugin.
- The condition is about one or more attributes from the
Warning
Ensure that your file evaluates to an empty result if the provided #var
value does not match the expected type.