[[!meta title="Subplot tutorial"]] NOTE: This tutorial is a little premature, as the software hasn't been written yet. The tutorial is based on an earlier prototype. Introduction ============================================================================= Subplot is a tool for **automated testing of acceptance criteria**. Acceptance tests define if a program (or other system) is accepable, from a user's or client's or employer's point of view. Basically, in a commercial consulting setting, acceptance tests determine if the client is willing to pay the bill for the software having been developed. Acceptance tests differ from other integration tests by looking at the software only from the user's point of view, not the developers'. Because of this, Subplot produces two things: a standalone document aimed at the user to describe the accepance tests, and a report from executing the automated tests. The document is meant to be a vehicle of communication between the various stakeholders in a project, for specifying in detail what the acceptance criteria are. The document both explains the criteria, and how they're verified, and lets the verification to be done automatically. Subplot generates the document, and a test program to execute the tests. The generated test program produces the test report. ```dot md [label="document source\n(Markdown)"]; md [shape=box]; bindings [label="bindings file\n(YAML)"]; bindings [shape=box]; impl [label="step implementations\n(Python, Rust, ...)"] impl [shape=box]; subplot [label="Subplot"]; subplot [shape=ellipse]; pdf [label="PDF document"] pdf [shape=box]; testprog [label="Test program\n(generated)"] testprog [shape=box]; report [label="Test report"] report [shape=box]; md -> subplot; bindings -> subplot; impl -> subplot; subplot -> pdf; subplot -> testprog; testprog -> report; ``` Example ============================================================================= Here is a quick example of using Subplot. Let's imagine you are in need of a new implementation of the **echo**(1) command line utility, and you want to commission the implementation. You need to specify acceptance tests for it. After negotations with the developers, and their project manager, and your own sales and test teams, you end up with the following requirements: * If echo is run without arguments, it should write a newline character, and exit with a zero exit code. * If echo is run with arguments, it should write each argument, separated by a space character, and then a newline character, and finally exit with a zero exit code. [Markdown]: https://en.wikipedia.org/wiki/Markdown These are specified for Subplot in a [Markdown][] document. A minimal one for echo might look like this: ~~~~markdown No arguments ============================================================================= Run `/bin/echo` without arguments. ```subplot when user runs echo without arguments then exit code is 0 and standard output contains a newline and standard error is empty ``` Hello, world ============================================================================= Run `/bin/echo` to produce the output "hello, world". ```subplot-scenario when user runs echo with arguments hello, world then exit code is 0 and standard output contains "hello, world" and standard error is empty ``` ~~~~ Acceptance tests in Subplot are expresses as **scenarios**, which roughly correspond to use cases and acceptance requirements. A scenario has an overall structure of given/when/then: * **given** some initial context (setup) * **when** some action is taken (the thing to test) * **then** the end result looks in a specific way (verify what happened) These are embedded in markdown using **fenced code blocks** marked as containing `subplot-scenario` text. Subplot extracts these and generates a test program to execute them. The scenario steps need to be formulated in a way that all stakeholders in the project understand. This is crucial. The scenario steps, and the Markdown document in general, is typically written by the software architect or analyst of the developing organization, in collaboration with the client, and approved by the client. Subplot is not magic artificial intelligence. Each scenario step needs to be implemented by programmers so that computers also understand them. Each step corresponds to a function, written in a programming language such as Python or Rust, and a YAML file **binds** the step to the function, using a pattern to capture relevant parts of the step. The pattern can be simple or a full PCRE regular expression. A binding file might look like this: ```yaml - when: user runs echo without arguments function: run_echo_without_args - when: user runs echo with arguments {args} function: run_echo_with_args - then: exit code is {exit_code} function: exit_code_is_zero - then: standard output contains a newline function: stdout_is_a_newline - then: standard output contains "(?P.*)" function: stdout_is_text - then: standard error is empty function: stderr_is_empty ``` This means that for a step saying 'then standard output contains "foo"', the Python function `stdout_is_text` is called and given the string `foo` as an argument. The developers on the project, in collaboration with the testing team, need to supply the bindings and the Python functions. Subplot produces the code that extracts the interesting parts of scenario steps, and calls your functions in the right order, plus any other scaffolding needed. Installing Subplot ============================================================================= To install Debian for Debian unstable, add the following to your APT sources lists: ``` deb http://ci-prod-controller.vm.liw.fi/debian unstable-ci main ``` Then run the following commands as root: ```sh apt update apt install subplot ``` To run Subplot from git, you need to clone the git repository and need Python 3, and the Markdown parser installed, as well as document typesetting tools, from Debian 10 (buster) or later: ```sh sudo apt update sudo apt install make git locales-all pandoc python3 \ python3-pandocfilters python3-commonmark-bkrs python3-yaml \ texlive-latex-base texlive-fonts-recommended \ graphviz librsvg2-bin git clone git://git.liw.fi/subplot cd subplot ``` Note that `subplot` contains sample versions of the files for the echo acceptance tests (`echo.md`, `echo.yaml`, `echo.py`, and `echo-prelude.py`) so that you don't need to type them in yourself. Producing a document ============================================================================= To produce a PDF document of the echo requirements, for you to review, the following commands are needed (prefix the commands with "./" if running from a git checkout): ```sh docgen --html echo.md docgen --pdf echo.md ``` `echo.md` is the markdown input file. `echo.yaml` is the bindings, and its name is inferred by `docgen`. `echo.pdf` and `echo.html` are the formatted documents. The input files are included in the `subplot` git repository. Running the tests ============================================================================= To generate the test program, and running it to produce the test report, the following commands are needed: ```sh codegen --run echo.md ``` The output of the last command is the test report: ``` OK: No arguments OK: Hello, world ``` `echo.yaml` and `echo.md` are again the bindings and markdown files. Also, `echo.py` has the Python functions, needed at runtime, though not referenced above. The names of `echo.py` and `echo.yaml` are inferred by `ftt-codegen`. Exercise ============================================================================= Your mission, should you choose to accept it, is to write a Subplot to write an acceptance test suite for simple uses of the **cp**(1) command. You can model it after the **echo**(1) one, and you get to specify the actual acceptance criteria yourself. For the exercise, it's enough for you to check that a regular file can be copied to another name in the same directory and that the end result is the same. You should try to get `docgen` produce a PDF of your subplot. You don't need to run the tests. Finally ============================================================================= It would great if you could give us, Lars and Daniel, feedback on Subplot and this tutorial. Questions for before you go through the tutorial (if we are interviewing you): * What kinds of testing do you already do? * Who are your usual stakeholders? * What sorts of things would you like to test better? Questions for after you go through the tutorial: * What do you think of Subplot as a concept? * What's good about Subplot? * What's less good about Subplot? * Did you have any problems following the tutorial? * Do you see yourself using Subplot in the future? * Do you see Subplot as filling a gap?