jkcfg / tekton Goto Github PK
View Code? Open in Web Editor NEWTypes and functions for building [Tekton](https://tekton.dev) configurations
License: Apache License 2.0
Types and functions for building [Tekton](https://tekton.dev) configurations
License: Apache License 2.0
I've recently received approval to contribute the TypeScript definitions I have authored internally at $DAYJOB
for Tekton in jkcfg to open source, and wanted to gauge interest in this style of API for this project. I could also see these type of functions making sense in a different package which uses jkcfg/tekton as a dependency.
For context, as an example, this is the file I wrote for generating Pipeline objects:
import { ResourceDeclarations, ResourceDeclaration } from 'lib/tekton/resource';
import {
ParameterSpecs,
Parameters,
ParameterValue,
ParameterSpec,
} from 'lib/tekton/param';
import { Workspaces, Workspace } from 'lib/tekton/workspace';
import { taskSpec, taskRef } from 'lib/tekton/task';
import { TaskOptions, TaskRef, Task } from 'lib/tekton/task';
import { KubernetesObject } from 'lib/models';
import { resource } from 'lib/tekton/common';
import { objToNamedObj, objToNameValue } from 'lib/util';
/**
* Resource models
*/
interface Pipeline extends KubernetesObject {
spec: PipelineSpec;
}
export interface PipelineSpec {
tasks: PipelineTaskSpec[];
resources?: ResourceDeclaration[];
params?: ParameterSpec[];
workspaces?: Workspace[];
}
/**
* Pipeline Task Resources
*/
interface PipelineTaskResource {
name: string;
resource: string;
}
interface PipelineTaskOutputResource extends PipelineTaskResource {}
interface PipelineTaskInputResource extends PipelineTaskResource {
// list of other pipeline tasks the resource has to come from
from?: string[];
}
/**
* Object format of PipelineTaskOutputResource[] for convenience
*/
interface PipelineTaskOutputResources {
[prop: string]: Omit<PipelineTaskOutputResource, 'name'>;
}
/**
* Object format of PipelineTaskInputResource[] for convenience
*/
interface PipelineTaskInputResources {
[prop: string]: Omit<PipelineTaskInputResource, 'name'>;
}
interface PipelineTaskConditionSpec {
conditionRef: string;
params?: ParameterValue[];
resources?: PipelineTaskInputResource[];
}
/**
* Object format of PipelineTaskConditionSpec for convenience
*/
interface PipelineTaskConditions {
[prop: string]: {
params?: Parameters;
resources?: PipelineTaskInputResources;
};
}
interface PipelineTaskSpec {
name: string;
taskRef?: TaskRef;
taskSpec?: Task['spec'];
resources?: {
inputs?: PipelineTaskInputResource[];
outputs?: PipelineTaskOutputResource[];
};
params?: ParameterValue[];
conditions?: PipelineTaskConditionSpec[];
retries?: number;
runAfter?: string[];
}
/**
* Abstract internal representation of what goes into PipelineTask
*/
interface PipelineTaskOptions {
name: string;
// simplify to string while keeping flexibility to use type for final interface
taskRef?: string;
taskSpec?: TaskOptions;
resources?: {
inputs?: PipelineTaskInputResources;
outputs?: PipelineTaskOutputResources;
};
params?: Parameters;
runAfter?: string[];
retries?: number;
conditions?: PipelineTaskConditions;
}
export interface PipelineOptions {
tasks: PipelineTaskOptions[];
/**
* Technically these are not the same in Pipelines, but shrug:
* https://github.com/tektoncd/pipeline/blob/release-v0.10.x/pkg/apis/pipeline/v1alpha2/pipeline_types.go#L170
*/
resources?: ResourceDeclarations;
params?: ParameterSpecs;
workspaces?: Workspaces;
}
/**
* Resource creation functions
*/
/**
* Creates PipelineTaskSpec from PipelineTaskOptions
* @param opts
*/
export const pipelineTask = (opts: PipelineTaskOptions): PipelineTaskSpec => {
const { name } = opts;
const spec: PipelineTaskSpec = { name };
if (opts.taskRef) spec.taskRef = taskRef(opts.taskRef);
if (opts.taskSpec) spec.taskSpec = taskSpec(opts.taskSpec);
if (opts.retries) spec.retries = opts.retries;
if (opts.runAfter) spec.runAfter = opts.runAfter;
if (opts.resources) spec.resources = {};
if (opts.resources?.inputs) {
spec.resources!.inputs = objToNamedObj<PipelineTaskInputResource>(
opts.resources.inputs
);
}
if (opts.resources?.outputs) {
spec.resources!.outputs = objToNamedObj<PipelineTaskOutputResource>(
opts.resources.outputs
);
}
if (opts.params) {
spec.params = objToNameValue(opts.params) as ParameterValue[];
}
if (opts.conditions) {
spec.conditions = (objToNamedObj(
opts.conditions
) as unknown) as PipelineTaskConditionSpec[];
}
return spec;
};
export const pipelineSpec = (opts: PipelineOptions): PipelineSpec => {
const spec: PipelineSpec = {
tasks: opts.tasks.map(t => pipelineTask(t)),
};
if (opts.resources) spec.resources = objToNamedObj(opts.resources);
if (opts.workspaces) spec.workspaces = objToNamedObj(opts.workspaces);
if (opts.params) spec.params = objToNamedObj(opts.params);
return spec;
};
/**
* Creates Pipeline object
* @param name
* @param opts
*/
export const pipeline = (name: string, opts: PipelineOptions): Pipeline =>
resource(name, 'Pipeline', pipelineSpec(opts));
In order to enable the more terse "jsonnet" style of defining Kubernetes named arrays as maps, I added additional types on top of the types which define the raw spec (those types which define the raw spec could probably be scrapped and the types from this project used in their place with a little work):
/**
* Pipeline Task Resources
*/
interface PipelineTaskResource {
name: string;
resource: string;
}
interface PipelineTaskOutputResource extends PipelineTaskResource {}
interface PipelineTaskInputResource extends PipelineTaskResource {
// list of other pipeline tasks the resource has to come from
from?: string[];
}
/**
* Object format of PipelineTaskOutputResource[] for convenience
*/
interface PipelineTaskOutputResources {
[prop: string]: Omit<PipelineTaskOutputResource, 'name'>;
}
/**
* Object format of PipelineTaskInputResource[] for convenience
*/
interface PipelineTaskInputResources {
[prop: string]: Omit<PipelineTaskInputResource, 'name'>;
}
By omitting the name and using the name as the key, I can now define things a lot more concisely, which brings joy to all, I think.
The really neat part, and I think where most of the value exists, is by using generics in TypeScript I can convert these "objectified" versions of arrays into arrays without losing any type checking:
if (opts.resources?.inputs) {
spec.resources!.inputs = objToNamedObj<PipelineTaskInputResource>(
opts.resources.inputs
);
}
if (opts.resources?.outputs) {
spec.resources!.outputs = objToNamedObj<PipelineTaskOutputResource>(
opts.resources.outputs
);
}
The generic utils for doing that are pretty stupidly simple, and might make sense to go into the @jkcfg/kubernetes
project?
/**
* Easily turn objects into arrays where the original key name is now `name:`
* and the value is now `value:`
*
* e.g, { foo: bar } => [{ name: foo, value: bar }]
*
* @param obj
*/
export const objToNameValue = (
obj: { [prop: string]: any },
valueKeyName = 'value'
) => Object.keys(obj).map(key => ({ name: key, [valueKeyName]: obj[key] }));
/**
* Easily turn objects into arrays of objects that have a `name` field based on
* their original key name. Contents of obj[key] are spread alongside name.
* @param obj
*/
export const objToNamedObj = <T = NamedObj>(obj: {
[prop: string]: any;
}): T[] => Object.keys(obj).map(key => ({ name: key, ...obj[key] }));
/**
* Turns ['name1', 'name2', 'name3' ] => [{ name: name1, ... }]
* @param arr
*/
export const arrToNamedObj = (arr: string[]) => arr.map(name => ({ name }));
Additionally, due to most Tekton resources being scheduled as part of a TriggerTemplate, I separate out the spec generation from the resource generation:
export const pipelineSpec = (opts: PipelineOptions): PipelineSpec => {
...
};
export const pipeline = (name: string, opts: PipelineOptions): Pipeline =>
resource(name, 'Pipeline', pipelineSpec(opts));
As an example of what this all looks like in context of generating a Tekton TriggerTemplate:
const prTriggerTemplate = triggerTemplate(
'pull-request',
{
gitrefafter: {
description: 'git ref pointing at HEAD of pull request',
},
gitrefbefore: {
description: 'git ref pointing at HEAD of base branch in pull request',
},
gitrepourl: {
description: 'git repository url',
},
pullrequesturl: {
description: 'pull request url',
},
gitbranch: {
description: 'head branch in pull request',
},
pullrequestnum: {
description: 'pull request number',
},
},
[
pipelineRun(k8sManifestsPresubmit.metadata?.name!, {
pipelineRef: k8sManifestsPresubmit.metadata?.name,
resources: {
source: {
resourceSpec: {
type: ResourceTypes.git,
params: objToNameValue({
revision: '$(params.gitbranch)',
url: '$(params.gitrepourl)',
}) as ResourceParameter[],
},
},
},
params: {
branch: '$(params.gitbranch)',
},
serviceAccountName: saName,
}),
pipelineRun(checkForLinkedIssues.metadata?.name! + '-$(uid)', {
// dont generate name so we can control the run name and produce URL
// reliably
generateName: false,
pipelineRef: checkForLinkedIssues.metadata?.name!,
serviceAccountName: saName,
params: {
url: pipelineRunLogsUrl(
checkForLinkedIssues.metadata?.name + '-$(uid)'
),
},
resources: { pr },
}),
].map(r => {
r.metadata!.labels = labels;
r.metadata!.annotations = annos;
return r;
})
);
Using all array based values would extend the length of a declaration like this quite a bit and add a bunch of boilerplate.
If there is any interest in these kind of patterns, I would be happy to work on refactoring my existing code to compose with what already exists in this repository and add some of our internal tasks/pipelines/etc as examples. Either way, I honestly just adore working with this project and am interested in your thoughts <3
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.