I am currently working with the AST in python. I take in a python file, generate its AST, modify it, and then recompile back to source code. I'm using a transformer that adds a getter to a class (I am using a visitor pattern with ast.NodeTransformer). Currently my code works as expected but does not preserve comments, which is my issue. Below is my code:
#visits nodes and generates getters or setters
def genGet(file,type,func):
global things
things['func'] = func
things['type'] = type
with open(file) as f:
code = f.read() #get the code
tree = ast.parse(code) #make the AST from the code
genTransformer().visit(tree) #lets generate getters or setters depending on type argument given in our transformer so the genTransformer function
source = meta.asttools.dump_python_source(tree) #recompile the modified ast to source code
newfile = "{}{}".format(file[:-3],"_mod.py")
print "attempting to write source code new file: {}".format(newfile) #tell everyone we will write our new source code to a file
outputfile = open(newfile,'w+')
outputfile.write(source) #write our new source code to a file
outputfile.close()
class genTransformer(ast.NodeTransformer):
...
I have done some research on lib2to3 which apparently can preserve comments but have not found anything as of yet that helps with my problem. For example, I found the code below but don't really understand it. It appears to preserve comments but not allow my modifications. I get a missing attribute error when it runs.
import lib2to3
from lib2to3.pgen2 import driver
from lib2to3 import pygram, pytree
import ast
def main():
filename = "%s" % ("exfunctions.py")
with open(filename) as f:
code = f.read()
drv = driver.Driver(pygram.python_grammar, pytree.convert)
tree = drv.parse_string(code, True)
# ast transfomer breaks if it is placed here
print str(tree)
return
I am having trouble finding a package or strategy to preserve comments whilst transforming an AST. Thus far my research has not helped me. What can I use that will allow me to modify an AST but also preserve the comments?
LibCST is a Python Concrete Syntax tree parser and toolkit which can be used to solve your problem. It provides a syntax tree looks like ast but with formatting info preserved. It also provides transformer pattern for tree modification.
https://github.com/Instagram/LibCST/
https://libcst.readthedocs.io/en/latest/index.html
With a NameTransformer like this, we can transform all names in source code to upper case: