apigeek-architect
v1.3.37
Published
Executable English for Software Teams
Downloads
249
Maintainers
Readme
Executable English for Agile Software Teams
ApiGeek:Architect is a testing, automation and orchestration tool for agile software teams.
Architect uses English as a scripting language to help simplify, standardise and socialise software deliverables.
We take a Business Driven Development (BDD) approach to designing, deploying and governing complex software systems.
Architect was designed to support a number of common use cases:
As an over-worked Developer, I want to build skeleton code using my favourite framework
As a QA Specialist, I want to generate executable BDD tests for APIs and web apps
As a Product Owner, I want to up-to-date document for various stakeholders
As an Enterprise Architect, I want to build, document and instrument my solutions
Tasks are written in English so that every stakeholder (including the CIO :-) can make sense of them.
I want to automate myself
We're all busy. We write scripts, or at we should. But our scripts are unreadable by our stakeholders - often even ourselves.
With Architect, you write your requirements in a simplified dialect of English - called Gherkin.
These "Features" contain executable English instructions - actions and assertions - that are easy to read, write and socialise.
Features are lists of related scenarios. It's the scenario's that do all the heavy lifting.
A scenario describes your pre-conditions, actions and outcomes in a way that is both human and machine friendly.
Architect can be invoked elegantly from an API, the command line, Mocha, your IDE or your DevOps workflow.
I want to learn Executable English
The BDD notation for a feature / scenario is:
Feature: Basic Examples
Scenario: An example
GIVEN some context
WHEN an action is performed
THEN an outcome is expected
The text following the keyword (GIVEN | WHEN | THEN) needs to match a phrase/pattern from a vocabulary.
This BDD notation is called "Gherkin". Gherkin is human and machine readable - business analysts, featureers, developers and robots can collaborate.
New features can be created using any simple text editor - where you'd use ApiGeek:Architect's extensible vocabulary to write them.
You can download pre-packaged vocabularies (called Dialects) and/or roll your own with simple Javascript.
I want to see a working example
To improve readability, the keyword AND can be used instead in place of the verbs above.
You can influence what ApiGeek:Architect understands using @dialect annotations.
@dialects=webapi
Feature: Verify that Google is accessible
Background: Google Scenarios
Given I am googling
Scenario: Request Google homepage - with redirects
Given I enable redirects
When I GET http://google.com
Then response code should be 200
Scenario: Request Google homepage - no redirect
Given I disable redirects
When I GET http://google.com
Then response code should be 3xx
I want to test-drive ApiGeek:Architect
ApiGeek:Architect is built using NodeJS.
If you're new to NodeJS, pre-read Installing Node.
You install ApiGeek:Architect as a system-wide CLI command:
$ npm install apigeek-architect -g
To run it simply type:
$ apigeek
However, it won't do much else until we provide some feature scenarios.
By default, ApiGeek:Architect looks for ".feature" files recursively, by default they're stored in the "./features/" directory.
I want to create my first Feature
To quickly create a few examples in the ./features folder, type:
$ apigeek --example
This will create the ./features folder and copy some simple examples.
It will also write your default configuration to ./apigeek.json
It won't damage if you run it again, except re-save your ./apigeek.json config.
Or, you can just create the ./features folder and a default "apigeek.json" without the examples:
$ apigeek --initialize
To execute your example ".feature" files, type:
$ apigeek
If something goes wrong, enable the built-in debugger.
$ export DEBUG=apigeek* $ apigeek
Now, the output is verbose and colour-coded to make your life easier.
To turn off debugging, type:
$ export DEBUG=
$ apigeek
I want to learn some vocabulary
ApiGeek:Architect ships with few default vocabularies - variables, files, web apis, web apps, etc.
Try out the included .feature examples - they cover a range of powerful use cases.
To discover what phrases exist in the vocabularies, type:
$ apigeek --knows
Let's create a trivial example of a hypothetical test case.
Scenario: Trivial Test
Given I am testing debug
When debug works
And I succeed
The steps are executed in sequence.
The GIVEN steps setup pre-conditions. The "Given I am $acting" phrase doesn't do much - except communicate intentions.
The "WHEN ... " steps do useful work that result in desirable outcomes. For example: writing a file, requesting a web page, etc.
In this example, we simply write a debug message to the console, so let's turn on debug output.
$ export DEBUG=apigeek*
You can adjust the logging scope - to see only Web API messages, use:
$ export DEBUG=apigeek:webapi
The "THEN ..." steps make assertions, that is they test that conditions are met. For example, you can use arbitrary Javascript if necessary:
Then I assert $some_javascript
The "I succeed", "I pass" always meet their conditions. The inverse "I fail" forces the scenario to abort and report it's failure.
I want to learn more about ApiGeek:Architect
For runtime options, type:
$ apigeek -h
We actively add capabilities and fix bugs. To upgrade to the latest version, type:
$ npm update apigeek-architect -g
For more information, please read:
I want to organise my work into folders
If your features are in a different location then use the "--features" option to locate them.
$ apigeek --archive ./my-archive --features ./my-features
These folders are not automatically created, they will throw an error if they're not found.
I want to re-use my features in other projects
ApiGeek:Architect was designed to support a declarative style so that features are portable.
Supplying a different "config" file for each environment allows features to be re-used across multiple environments.
For example, features can adapt to dev, QA and live environments - injecting hostnames, credentials, etc as required.
Most Dialects configure themselves automatically.
If yours doesn't then there is alternative - use {{mustache}} templates to modify statements prior to execution.
Given I login as {{scope.actor}}
In this way, your BDD features are neatly abstracted from your runtime configuration.
To specify a runtime configuration for your features, type:
$ apigeek --config ./my-context.json
By default, ApiGeek:Architect will try to load a configuration file called "apigeek.json" from your current directory.
If no file is found, then sensible default values are defined.
I want to do something before every scenario
Backgrounds are similar to scenarios, except they do not support annotations.
Any feature can contain a background, in which case the steps that carried out before each scenario.
Background: Authenticate
GIVEN I login
AND I use a valid client certificate
I want to know how it works
First, ApiGeek:Architect parses the command line and initializes the Dialect, Features and Engine components.
Next it loads the default dialects. These can be specified on using the APIGEEK_DIALECT environment variable. Dialects can also be specified using --dialect option and within Feature: definitions using the @dialect annotation.
Each Dialect instructs the parser (Yadda) to match a set of Gherkin phrases to their related function.
The Feature manager converts features and scenarios into executable units.
User defined variables are scoped at the feature-level meaning they are shared between scenarios within the same feature.
The --config file is used as the basis for internal context variables. They are scoped to each scenario.
These variables - such as web requests/responses - can be accessed using the "this." qualifier - with due caution.
Next, the Engine runs each feature using the built-in Mocha runner. Results are correlated and output according to your CLI options.
I want to add comments
It's useful to document your intentions or to prevent a statement from running, for example during development.
Simple, place a # before any line and it will be ignored by Architect.
# This is ignored by the parser
Scenario: Comments Example
Given I am using comments
# Then I fail
Then I succeed
Instead, you should use @skip or @todo before a Feature: or Scenario: definition.
An @bug scenario will pass normally (skipped) but fail when --debug is used.
@bug=something is broken
Scenario: A Bug
Given I am a bug
Then I fail
I want to automate everything
That is our goal too. We'll continue to address the needs of Enterprise Architects.
Competent software engineers can easily create "blueprints" that capture the patterns, templates and files used to build SDLC artefacts.
Then ApiGeek:Architect can re-use those blueprints to build customised websites, portals, Apps, APIs, mock servers, micro services and more.
Please share any custom Blueprints and Dialects so that ApiGeek:Architect becomes more useful for all of us.
If you need support to build or debug your community Blueprints or Dialects, please [email protected]
I want to license ApiGeek:Architect
This software is licensed under the Apache 2 license, quoted below.
Copyright 2015-2016 Lee Curtis & Troven 2009-2015
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
[http://www.apache.org/licenses/LICENSE-2.0]
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.