No more absolute links in R scripts

[alex-d13]

Currently we have many cases, where hard links are written in the R scripts. This should be fixed with a parameter or with another solution.

For example here: https://github.com/omnideconv/deconvBench/blob/dfba28afae790f64cba1dc19edd453cd35899fe5/pipeline/bin/computeSignaturesNF.R#L123

[grst]

If you are using nextflow, you should always write to the current working directory, and use publishDir to copy the files to their final destination. Writing absolute paths from a nextflow pipeline as an anti-pattern.

[alex-d13]

@grst You are right, we really did not follow the best practices well during the pipeline development :D

[alex-d13]

We still have a related issue with external files that should be accessed in R scripts, such as utils.R and optimal_normalization.csv. At the moment we use absolute paths to these files, which is not ideal. A solution could be with a new parameter, but its also not really pretty..

https://github.com/omnideconv/deconvBench/blob/2289c4007bf2eb32bfc9cef66f956f8115ba5773/pipeline/bin/analysisNF_mixed_simulations_real_dataset.R#L9 https://github.com/omnideconv/deconvBench/blob/2289c4007bf2eb32bfc9cef66f956f8115ba5773/pipeline/bin/analysisNF_mixed_simulations_real_dataset.R#L41

[grst]

I think it's fine to pass these files as an additional input channel. Another common pattern is to use something like ${baseDir}/bin/utils.R to retrieve files relative to the workflow root.

The cleanest solution, of course, would be to make an R package with helper functions and install it into the container.

[bernheder]

Solved by using nextflow baseDir variable #29