In metrumresearchgroup/mrgvalprep: Preprocessing for mrgvalidate

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = FALSE
)

Introduction

The mrgvalprep package contains helper functions for gathering, formatting, and preprocessing input data for mrgvalidate. The mrgvalidate package (as of the 2.0.0 release) is used to generate 7 specific documents that are necessary for the software validation process at Metrum Research Group. Those documents are:

• release-notes.docx
• validation-plan.docx
• testing-plan.docx
• requirements-specification.docx
• traceability-matrix.docx
• testing-results.docx
• validation-summary.docx

This vignette demonstrates some of the legacy functionality for using mrgvalprep to build validation docs from content stored in either Github issues or Googlesheets. Note that the current practice at Metrum Research Group is to store this data in YAML files inside the packages' version-controlled code repository instead. For a demonstration of that workflow, please see the "Basic Usage of mrgvalprep and mrgvalidate" vignette instead.

library(mrgvalprep)
library(mrgvalidate)

Full document creation

No matter how the Stories and Requirements are stored or parsed, the process of creating and formatting test outputs, as well as creating the final documents, does not differ from the process demonstrated in the "Basic Usage of mrgvalprep and mrgvalidate" vignette. For this reason, this vignette will only demonstrate the parsing of Stories and Requirements into a "specification" tibble.

Example 1: Stories and Requirements from Googlesheets

This example shows how to pull Stories and Requirements from Googlesheets.

Stories from Googlesheets

The validation process typically requires defining "user stories" that capture the promised functionality of the software, and mapping them to tests that verify this functionality. One way to do this with mrgvalidate is by keeping your Stories (and optionally accompanying Technical Requirements) in a Googlesheet that can be read in directly using read_spec_gsheets().

The format of these Googlesheets is described in the read_spec_gsheets() documentation. At a high-level, the Stories sheet must have the following columns:

• StoryId -- Unique identifier for the Story.
• StoryName -- Short human-readable name for the Story, to be printed in resulting docs.
• StoryDescription -- Short description of the Story, typically beginning with something like "As a user...".
• ProductRisk -- Label that conveys the level of product risk associated with this Story, typically "Low", "Medium", or "High".
• Either TestIds or RequirementIds -- Comma separated list of Id's for mapping Story to any associated tests or Technical Requirements.

Requirements

While Stories can be mapped directly to tests, it is often useful to instead map them to associated Technical Requirements, which are then mapped to specific tests. If you want to do this, you will need to include the RequirementIds column (and not the TestIds column) in your Stories Googlesheet, and then pass a second Googlesheet of Requirements with the following columns:

• RequirementId -- Unique identifier for the Requirement. These must match what is included in the RequirementIds column of the Stories Googlesheet.
• RequirementDescription -- Short description of the Requirement, to be printed in resulting docs.
• TestIds -- Comma separated list of Test Id's for mapping Requirement to any associated tests.

These Requirements will be included, bullet-pointed, under the relevant Stories in the resulting requirements-specification.docx document.

Note: If you do not want to use Requirements in your validation, simply pass a Stories sheet containing a TestIds column and not a RequirementIds column.

Calling read_spec_gsheets() will pull the data in your sheet(s) into a tibble that can be passed directly to mrgvalidate::create_package_docs().

spec_df <- read_spec_gsheets(
  ss_stories = "1HgsxL4qfYK-wjB-nloMilQiuBSLV6FZq_h2ToB6QNlI", # the sheet with Stories
  ss_req = "1SnyUzxVDUUJFtMGEi2x4zJE0iB2JDV1Np-ivhaUkODk"      # the (optional) sheet with Requirements
)

Example 2: An R package on Github, with Stories from Github Issues {#example2}

This example shows how to validate an R package that is:

• stored in a Github repository ("repo")
• uses R's testthat framework
• has Stories defined in Github issues

This workflow will clone the repo locally, run its test suite and save the results, and fetch Stories and Requirements from Issues on the Github repo.

Initial setup

Before proceeding, you will need to have ghpm installed on your system in order to interact with Github. This is an internal Metrum package that handles calls to the Github API. You can try library(ghpm) if you're not sure if you have it. If you do not, install it with:

devtools::install_github("metrumresearchgroup/ghpm")

Next, you will need to make sure you have a valid token assigned to the relevant environment variable so that ghpm can access the Github API. Depending on whether the package you are validating is on Github Enterprise or public Github, you will either need to set GHE_PAT or GITHUB_PAT respectively. If you do not have a token please read an article like this for instuctions on getting one set up. Then use the following code to set the variable (substituting GHE_PAT if appropriate).

Sys.setenv(GITHUB_PAT="your-token") # "your-token" will be the alphanumeric OAuth token you get from Github

Test outputs

While you can call parse_testthat_list_reporter() directly, as shown in the "Basic Usage" vignette, if you are validating an R package on Github it may be easier to use the validate_tests() wrapper. This does several things:

• Clones the repo of the package being validated from Github at a specified tag
• Installs the package in a temp directory
• Runs the package's test suite with devtools::test(reporter = testthat::ListReporter)
• Passes that output to parse_testthat_list_reporter() and writes the resulting tibble to a .csv
• Calls get_sys_info() to capture relevant system and user information, writing that to a .json

You will then put the resulting .csv and .json files into a directory that will be passed to the auto_test_dir argument of mrgvalidate::create_package_docs().

validate_tests(
  org = "metrumresearchgroup",        # Github organization containing repo that will be validated
  repo = "mrgvalidatetestreference",  # repo for fake package used by mrgvalprep test suite
  version = "0.6.0",                  # the tag that will be pulled for testing
  out_file = "test_outputs/test_res", # basename for output .csv and .json files
  set_id_to_name = TRUE               # pass this only if _not_ using Test Id's (described below)
)

This call will write the testthat output to test_res.csv and the system info (including the commit hash of the git tag you pass to version) to test_res.json, both in the test_outputs directory.

Identifying tests in code

To map tests to Stories, each test must be uniquely identified. The preferred method is to include Test Id's in the Test Name within your code, as described in the "Basic Usage" vignette. However, it is also possible to match on the full Test Name (the full string passed as the first argument to test_that()). For example:

test_that("feature X works correctly", {
  # test code
})

In this case, feature X works correctly is the Test Name which will be passed through as a Test Id, if you pass validate_tests(..., set_id_to_name = TRUE) as shown above. Your Stories can then reference this full string, as discussed in the next section. Note that, for mapping tests to Stories, you must either use Test Id's or the full Test Names. It is not possible to mix between them within a test suite.

A few notes of caution on using full Test Names:

• All tests must have a unique name specified as the first arguement to the test_that() function (as in, no two tests have the same name).
• If using the describe()/it() syntax in testthat, the Test Name will be the concatenation of the string passed to the outer describe() and the inner it().
• If you are using Test Names in your issues then, if you change the names of your tests in future versions, you will have to go back and change the names in the issues as well (as described in the "Issue format" section below). This is obviously toilsome and is one of the primary motivations for using Test Id's instead.

To reiterate, for these reasons and others, Test Id's is the preferred method for mapping test outputs to Stories.

Stories from Github issues

You can define your Stories in Github issues, attached to a milestone, and mrgvalprep will pull them into a tibble that can be passed to mrgvalidate. However, doing this relies on the input data being structured in a predictable way:

Milestone

All issues relevant to the release you are validating must be associated with the same milestone. This may be named the same as the tag for the release commit, but it does not need to be.

Note that this works well for validating a change set, but makes it difficult to validate the full functionality of the package (because all functionality would need to be represented in an issue on this milestone). If you need to validate the full functionality in single set of docs, consider using YAML files (shown in the "Basic Usage" vignette) or Googlesheets (shown in Example 1 above).

Issue format

The issue parser for creating validation documents requires a very specific format for all issues that are being validated (i.e. associated with the milestone discussed above).

• Summary section
• There must be a top-level heading entitled # Summary
• This should contain a "user story" describing the functionality
• Tests section
• There must be a top-level heading entitled # Tests
• This should contain a bulleted list of all the tests that are relevant to this functionality. This list can take two forms:
  • Test Id's -- as described in Example 1, you can list only the Test Id's (like XXX-XXX-123) here, as long as those Test Id's are contained in the .csv file created by validate_tests(). See Example 1 above for how to included Test Id's in your test suite code.
  • Test Names -- alternatively, you can pass the full text of the Test Name (i.e. the string passed as the first arguement to the test_that() function in your test code). As mentioned above, this method is not recommended because it relies on exact full string matching and is therefore more brittle than using Test Id's. If you use this method, you will need to pass validate_tests(..., set_id_to_name = TRUE) as shown above.
• Note that if an issue has no tests associated with it, you still need a top-level heading entitled # Tests but it should contain only some brief text about why tests are not necessary (with no leading bullets).
• Risk
• All issues must have a label that conveys the level of product risk associated with this functionality.
• Label must be formatted like risk: ____

There should be no other top-level headings. Examples of correctly formatted issues can be seen in the test reference repo used by the mrgvalprep test suite.

Pulling the issues

Once you have your issues formatted as described in the previous section, you can pull them into a tibble with parse_github_issues().

stories_df <- parse_github_issues(
  org = "metrumresearchgroup",        # Github organization containing repo that will be validated
  repo = "mrgvalidatetestreference",  # repo for fake package used by mrgvalprep test suite
  mile = "v0.6.0",                    # the name of the milestone in Github
  prefix = "PKG"                      # acronym of 3 letters signifying the associated package to use for Story IDs
)

Generating the docs

Now that you have test outputs and a tibble of Stories, you are ready to generate your validation documents. This is done with mrgvalidate::create_package_docs(), as shown in the "Basic Usage" vignette.

You will have .csv and .json files of test output, from your validate_tests() call above, in the test_outputs directory. Simply pass the path to that directory, along with your tibble of Stories, into mrgvalidate::create_package_docs().

metrumresearchgroup/mrgvalprep documentation built on Dec. 10, 2022, 3:17 a.m.