appsilon / shiny.benchmark Goto Github PK

Fix the current status of the package to make it easier to maintain.

It seems that without these libraries the code just freeze

Right now git2r is not needed since most of the git functions are done by system("code") and we only need simple functions. With time we may need some more complex solutions and then we can reconsider using git2r. Right now it is only increasing complexity.

Just use the name to user choose

For some reason the last test in the Cypress list of tests does not works. It does not appear in the final list of time measurements.

Let's use shiny.beanchmark instead of shiny.performance

Currently only Cypress tests are available. However, it is possible to create tests using R purely through shinytest2.

Functions like system(cp -r folder1 folder2) are more likely to work on linux. Try to use file.copy and file.symlink instead

If the current branch has uncommitted work, git checkout branch_name will fail. The message is not too informative so it is better to test it in advance to show the user a nice message about the problem.

We do not want to compare performance of all the tests (normally). We should be able to identify which tests are needed for the performance check and which ones are not.

Break the functions into smaller ones and add tests when necessary

The beta version has some lint issues. Fix it using linter package and keep it organized for the further implementions

Try to use the word benchmark to create the idea of replications and comparison

Add PR template based on Appsilon standard

Add information such as:

• How to install the package
• What are the mandatory setup
• How to install Cypress in your machine
• How to use the package

stop("You must provide 1 or {n_commits} paths for {obj_name}") should be stop(glue("You must provide 1 or {n_commits} paths for {obj_name}"))

Failing tests: https://github.com/Appsilon/shiny.benchmark/actions/runs/3361979812/jobs/5573090092
PR comments: #79 (comment)

It would be great to clean the files after executing the tests. Currently, if n_rep is greater than 5, it freezes my computer.

Run the tests on different OS:

• Linux (Ubuntu)
• Windows
• Mac

Right now it is impossible to run the tests in silent mode. It will be better to have something like options(shiny.performance.logs = "debug") info, error and so on...

Provide an example to show basic functionalities of the package. Do not run it during CRAN checks.

The package already have several functions but do not have any tests. Tests are important to make sure that new developed functionalities are not messing the previous ones.

Right now the 1/n message basically does not appear

Test can fail because of different problems e.g.:

• not able to find the desired commit
• not able to run renv::restore
• not able to find the set of tests
• permission issues

We should be able to inform the user the source of the error

It should be a Quarto template. The main idea is to have a wrapper like shiny.performance::create_report("path_to_report.md") and shiny.performance::render_report("path_to_report.md"). This way, the user can edit the path_to_report.md file instead of just observing the result.

Double check documentation and be sure it is clear enough for users

It is a little bit trick to test the functionalities of this package since we need an application and several branches with different setups. Therefore, it would be great to create a script able to setup an example application and all the characteristics desired (renv/no renv, single file/multiple files, no repetition/several repetitions, ...).

Implement function:

• check_renv: is renv is use?

Sent an error message if renv is not in use in a specific branch

Update repo name so GitHub actions can run the tests properly

Animals:

List of the fastest animals in the earth:
https://en.wikipedia.org/wiki/Fastest_animals#:~:text=10%20References-,Fastest%20organism,conflicting%20reports%20of%20recorded%20speeds.

Peregrine Falcon - Fastest animal on earth
Cheetah - Fastest land animal on earth

Package name:

shiny.beanchmark
shiny.performance

Color:
Red (#FF0000) is the fastest color in a vacuum space
Violet (#8F00FF) is the slowest color in a vacuum space

f <- colorRampPalette(colors = c("#FF0000", "#8F00FF"))
f(3)[2]

Allow user to provide:

• single test file (run the same set of test in all branches)
• multiple test files as input (run specific tests for each branch)
• multiple test files stored in the specific commit (tests/ folder)

When we run the tests only once, we do not have any idea about the variability of the execution times. We should be able to run the tests as many times as requested.

Avoid double messages related to switching branches

We do not want to add/remove files from the current git folder. In this case, it is preferable to run all the tests in a temporary folder. It will work unless we have hard coded paths in the application.

Be sure parameters are intuitive

Add GitHub actions to run lint and package check

Add requested packages and fix meta information