VarWizard is a language model-based tool to help generate more meaningful variable names in source code than the original version. It supports 10 programming languages: c, cpp, java, javascript, go, python, php, c_sharp, ruby, rust.
An overview of VarWizard is in the following figure. VarWizard automatically extracts all variable names in a code snippet, and then it generates candidates for them.
You can easily install this package by the command
pip install varwizardYou can use VarWizard by running the command
varwizard [--model-name {bloom-560m, codet5-base} (default: bloom-560m)]
--input INPUT --lang {c,cpp,java,php,go,javascript,ruby,rust,python,c_sharp}
[--output-path OUTPUT_PATH (default: None)] [--max-input-len MAX_INPUT_LEN (default: 400)]
[--device DEVICE (default: cpu)] [--penalty-alpha PENALTY_ALPHA (default: 0.6)] [--top-k TOP_K (default: 4)] [--max-new-tokens MAX_NEW_TOKENS (default: 100)]Details for each argument can be found by
varwizard --help
Another way is to use Python Apis. Here is a simple example in this way.
from varwizard import VarWizard
model = VarWizard(model_name = "bloom-560m")
code = """
function chunkData(str, chunk) {
var chunk = [];
var length = str.length;
var i = 0;
for (; i < length; i += chunk) {
if (i + chunk < length) {
chunk.push(str.substring(i, i + chunk));
} else {
chunk.push(str.substring(i, length));
}
}
return chunk;
}
"""
print(model.make_new_code(code, 'javascript', device = 'cuda:0'))VarWizard produces the output
function chunkData(data, length) {
var result = [];
var lengthOf = data.length;
var pos = 0;
for (; pos < lengthOf; pos += length) {
if (pos + length < lengthOf) {
result.push(data.substring(pos, pos + length));
} else {
result.push(data.substring(pos, lengthOf));
}
}
return result;
} You can play at the link: https://varwizard.loca.lt. At the first time to access, you may need to enter: 4.193.50.237
There are some examples for VarWizard's usage. You can navigate to the folder examples and then go to any subfolder to run the script.