C++ Linting

[joey-kilgore]

I've been recently trying to understand and search through the CARLsim codebase for the parameters used in the COBA mode calculations. I have noticed several instances where coding style varies drastically, and sometimes to the detriment of my ability to search the codebase using regex (or even just a simple ctrl+f). As an example

CARLsim6/carlsim/kernel/src/snn_cpu_module.cpp

Line 1819 in d527c55

runtimeData[netId].gAMPA [postNId] += change * mulSynFast[mulIndex]; // scale by some factor

has a space between gAMPA and [postNId] which makes some searching challenging even though normally there shouldn't be any space between an array and the indexing.
Earlier in that file we can see other issues with curly brace formatting

CARLsim6/carlsim/kernel/src/snn_cpu_module.cpp

Lines 1330 to 1336 in d527c55

	if (dPar_old.delay_length >= 0) {
	dPar_old.delay_length--; // decrement
	dPar_old.delay_index_start = 0; // reset entry
	}
	else {
	KERNEL_ERROR("Post-synaptic delay was not sorted correctly pre_id=%d, offset=%d", lNIdPre, offset);
	}

But later in that same method we see

CARLsim6/carlsim/kernel/src/snn_cpu_module.cpp

Lines 1362 to 1369 in d527c55

	if (offset == n) {
	setDelay(connInfo.nSrc, connInfo.nDest, connInfo.delay); // update delay cache
	}
	else
	{
	printf("WARNING: skipping setDelay (offset!=n): pre=%d, post=%d delay=%d\n",
	connInfo.nSrc, connInfo.nDest, connInfo.delay);
	}

And later in that same file

CARLsim6/carlsim/kernel/src/snn_cpu_module.cpp

Lines 2740 to 2746 in d527c55

	if (lGrpId == ALL) {
	lengthN = networkConfigs[netId].numNAssigned;
	posN = 0;
	} else {
	lengthN = groupConfigs[netId][lGrpId].numN;
	posN = groupConfigs[netId][lGrpId].lStartN;
	}

These are clearly small problems with the code base, but they create challenges for code readability and sometimes making understanding the code difficult (and challenges the ease of searching and traversing the codebase).

I'm also sure that there is talk about how many pieces of commented out code are covered throughout the entire repository. While these chunks of code appear to have been added long ago, there are some that are very newly added. I know they are additional debug statements, but CARLsim has a specific KERNEL_DEBUG for this (see d527c55) which I think is wonderful and helps keep the code clean.

I would really love to see some form of linter being implemented in the repository to clean it automatically. CARLsim is such a useful piece of software, and quite sprawling with features, that manually linting would be impractical, and there are plenty of tools that exist to set this up. I don't have much experience with linting C++ but I've linked one below that seems good (I've set it up for python projects, and I'm assuming the process is similar with just some slight nuances for the specific linter being used). It appears to have samples setup for using GitHub Actions as well, so hopefully that can simplify the transition as well!

https://github.com/cpp-linter