Skip to content

Development

Setup development environment

git clone https://github.com/mondeja/http-request-codegen.git
python3 -m virtualenv venv
. venv/bin/activate
python3 -m pip install .[dev]
pre-commit install
git clone https://github.com/mondeja/http-request-codegen.git
python3 -m virtualenv venv
venv\Scripts\activate.bat
python3 -m pip install .[dev]
pre-commit install

Development commands

Test

pytest -sv
pytest -s --cov=http_request_codegen --cov-config=setup.cfg --cov-report=html
pytest -svv --doctest-modules http_request_codegen

Lint

pre-commit run --all-files

Developing implementations

To develop an HTTP method function for a library or a program, you need to take in accounts all parameters described in generate_http_request_code function, but not the randomized values passed in parameters argument, because http_request_codegen provides functions that can handle these.

Implementation arguments

Each implementation function must contain the following arguments, which are passed from the API function generate_http_request_code, so it's recommended that you familiarize yourself with the arguments of that function before continuing with this guide because they are well documented there:

  • url: unique positional argument of the function, represents the target URL of the request.
  • headers: dictionary of headers.
  • parameters: list of parameter data objects.
  • files: dictionary of files, only passed to POST requests, so this should not be defined as argument is the function implementation's name is different to post.
  • wrap: maximum anchor of the rendered code snippet.
  • indent: indentation used in the rendered code snippet.
  • quote_char: string quotation character.
  • setup : code snippet prepended to generated request output.
  • teardown : code snippet appended at the end of the generated request output.
  • oneline: if enabled, render the code snippet in one line.
  • seed: seed used generating random fake values of parameters.
  • locale: locale used by faker library to localize the faked random values for parameters.

Method singularities

POST

    Most POST methods implementations render their code snippets different,
    depending on *Content-Types* header, including by default some of the
    most used *Content-Types* header related behaviours:

    - The default behavior, even if you don't specify it explicitly in the
        *Content-Type* header is the generation of an
        `application/x-www-form-urlencoded` encoded request.
    - If you want to generate a ``multipart/form-data`` encoded request,
        you need to specify the files to sent using the ``files`` argument.
    - If you specifies the *Content-Type* header `application/json`, the
        parameters sent will be adjusted according to the JSON encoded POST
        request.
    - If you specifies the *Content-Type* header `text/plain`, you can only
        send one parameter and it will be adjusted accordingly following
        the implementation.

One line wrapping behaviour

The first thing to take in account (and the most complicated one) is the behaviour of wrapping (wrap argument) rendering as if oneline=True is passed. The question is: can a snippet of code be outputted in one line if the estimated length of the request is lower than wrap argument value?

For example, Python requests can be rendered using this kind of code, in one line:

import requests

req = requests.get('https://github.com/mondeja/http-request-codegen')

...or using multiple lines (wrap is lower than expected length):

import requests

req = requests.get(
  'https://github.com/mondeja/http-request-codegen'
)

Of course, this also affects parameters, headers and kwargs:

import requests

req = requests.get('localhost', params={'foo': 'bar'}, headers={'foo': 'bar'})

...which can be outputted in multiple lines:

import requests

req = requests.get(
    'localhost',
    params={'foo': 'bar'},
    headers={'foo': 'bar'}
)

Since this behaviour can depend both oneline and wrap arguments, the recommended way of implement this is to calculate the length of the expected request inside the code snippet, and, if it is greater or equal to wrap argument, must be rendered as if oneline=True.

Tip

You can see an example of this type of implementation at http_request_codegen.generators.python.requests::get function.

But other implementations could be rendered in multiples lines regardless the wrap argument value.

For example, the Javascript fetch API implementation will output always a multiline code snippet (unless oneline=True is explicitly defined), because the Javascript Promises writing in single line is not a common syntax and there is a little chance that the generated request could not be wrapped given the default wrap value (80 in this case). The minimum reasonable possible code snippet in one line for Javascript fetch API implementation would be:

fetch('localhost').then(function(response) {}).catch(function(error) {console.error(error)});

...which exceeds the default wrap value length (80). In such type of cases, there is no need of calculate the expected generated code snippet request length before build their output.

Tip

You can see an example of this type of implementation at http_request_codegen.generators.javascript.fetch::get function.

In the first case, you need to iterate over parameters, headers and kwargs arguments to compute the expected length, then compare the expected length with wrap argument value and, if it reaches it, define an internal oneline=True like behaviour. In the second, you can assume that the generated code is multiline unless oneline=True is explicitly defined as argument.

Randomizing values

The library provides the functions lazy_name_by_parameter and lazy_value_by_parameter which returns the name and the value of a parameter given a parameter dictionary specification. These must be used to randomize parameters in a unified way across implementations as described in generate_http_request_code function documentation.

Language/platform utilities

You can create an _utils.py module inside a language or platform package to store utilities that could help in the process of building the code snippet, like:

  • Define default indentation for the language/platform (indent argument).
  • Define default wrapping length value (wrap argument).
  • Define default quotation character/s (quote_char argument).
  • Escape quotes of values (according to given quote_char argument).
  • Create greater level functions of code generation for the language/platform, such as string definitions with wrapping behaviour, dictionary definitions...

Tip

See current _utils.py modules of generators packages as reference.

Creating test cases

Use the script scripts/create-impl-test-cases.py to create possible generated code snippets cases accordingly to combination of arguments. This will help you developing implementations because saves you the need of execute every possible combination of arguments. Use it as follows:

rm -rf cases && python3 scripts/create-impl-test-cases.py \
  --language python \
  --implementation requests \
  --method GET \
  --directory cases

Previous command will create a cases/ directory with a lot of code snippets generated, given the combinations described in tests/combinations.py.

When you will have manually revised that all code snippets are generated correctly, you can create a test for the implementation at tests/test_generators/test_<lang>/test_<impl>/test_<impl>.py, placing the cases/ directory at tests/test_generators/test_<lang>/test_<impl>/<METHOD>.

For example, for Python requests GET method, the test module would be tests/test_generators/test_python/test_requests/test_requests.py and the cases/ directory would be placed at tests/test_generators/test_python/test_requests/GET/.

Tip

You can use an already implemented test module as a reference to write the one for the implementation.