google / synergyfinderengineered Goto Github PK

This is not an officially supported Google product.

This project is a fork of the 'synergyfinder' project, available in github at https://github.com/hly89/synergyfinder, and also available in BioConductor. The code here was originally offered as a patch to the main project, with some code clean-up, bug fixes, and unit tests. The patch proved difficult to complete (in fact a fork was specifically suggested by the institution that maintains the original project), so the code is offered here instead.

SUMMARY

The biggest part of this change is 'refactoring' of PlotSynergy according to (my humble interpretation of) software engineering best practices. The overall logic of the original code is maintained, but is now expressed as a set of related functions with the following properties:

• The new functions are each small, with relatively few local variables.
• The new functions are documented with comments.
• The new functions are tested with unit tests.
• The new tests provide ways to see specific expected inputs and outputs at a code level (rather than just an output level), and to ensure that functions fulfill their interface contracts even if reimplemented. For example, a new function that rounds a value up to the next multiple of 10 has a test with a few input values. If some other method of rounding is chosen in the future, it may be deemed correct if it passes the same tests that the current version passes. If it does not pass, it may still be correct in some sense; but the tests will (by breaking) explicitly highlight the behavior change so that callers of the function may be adjusted as needed.

Much logic that was replicated between '2D' and 'all', or between '3D' and 'all', is now consolidated and shared. For example, the only major difference in 2D plots between the '2D' and 'all' cases was a single layout directive issued before any part of the plot was drawn; yet all of the plotting logic was duplicated. Now, small functions retain the logic that was held in common, and the layout directive is issued by the caller before the functions are invoked.

There are some changes to the vignettes. They are not all technical. I have gone so far as to suggest some modifications to English content. It should be reviewed very carefully to ensure continued scientific correctness. The technical changes permit the following:

• All of the vignettes now execute end-to-end and uninterrupted when RStudio's 'run all' button is clicked.
• The vignettes fully execute and generate their PDF output when RStudio's 'Compile PDF' button is clicked.
• RStudio's 'Check' feature inspects the entire package and completes successfully.

CHANGE DETAILS

1. Updates to LICENSE. I fixed an apparent copy/paste error from the R DESCRIPTION template, and added my employer to the list of copyright holders.
2. Various updates to DESCRIPTION.
   2.1. Update 'imports' to include 'SpatialExtremes' instead of 'kriging'.
   2.2. My name is added to the 'Author' list in this file.
   2.3. Long lines are wrapped, for easier readability.
   2.4. Reduce the required version of ggplot2.
3. Eliminate all mention of 'kriging' from NAMESPACE.
4. Make moderate adjustments to Plot_dose_response.R.
   4.1. Adjust semantics for displaying a plot or saving it to a file.
   ..4.1.1. Eliminate post-loop plot 'replay'.
   ..4.1.2. Use dev.new() only when not saving to a file.
   ..4.1.3. Use dev.off() only when saving to a file.
   4.2. Use explicit package namespace prefixes for calls to functions in 'ggplot2' and 'grid'.
   4.3. Remove some commented-out code, clean up some whitespace, reflow some comments.
5. Perform a near-rewrite refactoring of Plot_synergy.R.
   5.1. Adjust semantics of displaying/saving plots, similarly to Plot_dose_response.R. Furthermore, update output filenames to include 'type' ('2D', '3D', or 'all').
   5.2. Use numerous tightly-scoped, commented functions for individual pieces of functionality.
   ..5.2.1. On the simple side, this includes some one-line specialty array-slicing implementations, and a function to take an argument's absolute value and round it to the next highest multiple of 10.
   ..5.2.2. On the more complicated side, this includes use of helper functions for 2D and 3D plots. The upshot is that the plotting functionality is now shared with the 'plot all' feature, instead of replicated for it. This eliminates significant code duplication.
   5.3. Add comprehensive unit tests (in test_Plot_synergy.R) for Plot_synergy.R helper functions that determine the data to be included in plots. The plot generation functions themselves are not unit-tested, but retain very little non-trivial logic.
   5.4. Multiple adjustments to whitespace and comment flow. API documentation has been added for PlotSynergy parameters that were previously undocumented.
   5.5. Explicit 'for' loops have been eliminated in favor of 'lapply'.
   5.6. PlotSynergy is now defined to return NULL, but with a side effect of either saving a file or displaying a plot.
6. Several updates to the vignettes in synergyfinder.Rnw:
   6.1. A few minor English tweaks either adjust grammar for clarity/rigor or satisfy TeX's 'badness' warnings. There are various whitespace changes as well.
   6.2. Uses of 'h!' that TeX warned it was rewriting as 'ht!' are now explicitly 'ht!'.
   6.3. File names of dose.reponse PDF inclusions now actually match (including with '.pdf' suffix) the files that are meant to be included. Quotes are removed because they were being parsed as part of the file names.
   6.4. Examples for PlotSynergy() now explicitly execute for every combination of 'type' and 'save.file'.
   6.5. The filenames of PDF output to be included in the vignette document are adjusted to include 'type' (which, as noted above, is a newly-added constituent of the filename in this change).