quarto_render(): invisibly return paths of all rendered files on success

[alexpghayes]

This would make it much nicer to use quarto_render() with @wlandau's targets package to produce large volumes of parameterized reports. Currently it's a bit painful to produce parameterized reports when there are multiple output formats and you would like to render to them all at once in a targets pipeline.

If there some other easy way to programmatically access the names/paths of rendered files, that would also be great.

[wlandau]

I added Quarto integration to https://github.com/ropensci/tarchetypes through new functions tar_quarto() and tar_quarto_rep(). The major challenge was to identify all the important files so that targets can tell when to re-render a project or document. The tar_quarto_files() function digs through the output of quarto::quarto_inspect() to figure this out, but I am 100% sure if I am doing it right. For standalone documents, tar_quarto_files() looks at quarto_inspect()$formats$html$pandoc[["output-file"]], etc. For projects, it looks at quarto_inspect()$files quarto_inspect()$config$project[["output-dir"]]. Am I missing anything? Are there edge cases I am not capturing, or variations on the projects or documents where the output of quarto_inspect() looks different?

As for large volumes of parameterized reports, tarchetypes::tar_quarto_rep() sets up a pipeline to render a single source document repeatedly over a grid of input parameters. tar_quarto_rep() detects the user's choice of output format from quarto_inspect(), and then it uses the format to create an output file name for each set of parameters in the grid. tar_quarto_rep() uses the output_file argument of quarto_render() to enforce the new file name, but this breaks for multiple formats because output_file cannot be of length greater than 1. Would you be open to allowing output_file and output_format specify more than one file output in quarto_render()?

[wlandau]

In fact, it would be really nice if quarto_inspect() could accept an output_format argument (which could be a vector) so Quarto could tell tarchetypes exactly which files it will generate (files which targets should track for changes). Came up in ropensci/tarchetypes#102.

[wlandau]

Another edge case came up at ropensci/tarchetypes#107. If there is a simple _quarto.yml file, then quarto_inspect("single_report.qmd") does not show the output directory where quarto_render("single_report.qmd") will store single_report.html. For integration with targets, it would be extremely helpful if quarto_inspect() could straightforwardly list all the output files, with directory paths included, that will be generated by quarto_render(). Otherwise, tarchetypes would need to emulate the output file policies of Quarto, which could change if Quarto changes.

[cderv]

Thanks for your feedback. I wonder if this kind of feature shouldn't be more generic that just the R package. I believe this discussion could be related quarto-dev/quarto-cli#2372
You may want to contribute to the discussion there.

Probably quarto inspect or a variant should try to handle that. If it is not in the scope of the CLI, we can see what could be done at the R level after running quarto.

[wlandau]

Sure, sounds good. I opened quarto-dev/quarto-cli#2378 and referenced this thread.

[cderv]

I opened an issue directly to track this more clearly in quarto-cli - this needs to be done there IMO

• [FR] return all output files from quarto inspect quarto-cli#8397

[wlandau]

Thank you, this would really help!