CoilSnake passes an explicit list of all CCScript files in a project to the CCScript compiler extension module code:
|
# Compile scripts using CCScript |
|
script_filenames = [os.path.join(project_path, "ccscript", x) |
|
for x in os.listdir(os.path.join(project_path, "ccscript")) |
|
if x.lower().endswith('.ccs')] |
|
|
|
if script_filenames: |
|
log.info("Compiling CCScript") |
|
if not ccscript_offset: |
|
ccscript_offset = "F10000" |
|
elif type(ccscript_offset) == int: |
|
ccscript_offset = "{:x}".format(ccscript_offset) |
The use of os.listdir appears to mean that only .ccs files in the top-level ccscript directory, and not any of its subdirectories, will be added to the build. This has caused pain as people tried to put dialogue and important data, rather than just macros, in subfolders; explicit imports of these files are needed for them to be built.
It's also a potential source for non-determinism in the build -- the list of path entries returned "is in arbitrary order," rather than something predictable like alphabetical order. I think that means you could get a different result if you made a .zip of a project and then sent it to others and the .zip was extracted in alphabetical order. It would be best to squash that as well by sorting all the paths.
(I was curious how Visual CCScript handled this. It seems like .pkproj files contain an explicit list of all files to compile. It might be fair to label this CoilSnake "problem" as being inherited from the convention in the days of Visual CCScript -- notice how Zorkbound doesn't include anything from the Core folder in the project file itself -- but it's still kind of annoying nowadays.)
CoilSnake passes an explicit list of all CCScript files in a project to the CCScript compiler extension module code:
CoilSnake/coilsnake/ui/common.py
Lines 113 to 123 in d806fa6
The use of
os.listdirappears to mean that only .ccs files in the top-levelccscriptdirectory, and not any of its subdirectories, will be added to the build. This has caused pain as people tried to put dialogue and important data, rather than just macros, in subfolders; explicitimports of these files are needed for them to be built.It's also a potential source for non-determinism in the build -- the list of path entries returned "is in arbitrary order," rather than something predictable like alphabetical order. I think that means you could get a different result if you made a .zip of a project and then sent it to others and the .zip was extracted in alphabetical order. It would be best to squash that as well by sorting all the paths.
(I was curious how Visual CCScript handled this. It seems like .pkproj files contain an explicit list of all files to compile. It might be fair to label this CoilSnake "problem" as being inherited from the convention in the days of Visual CCScript -- notice how Zorkbound doesn't include anything from the
Corefolder in the project file itself -- but it's still kind of annoying nowadays.)