cstq

A very simple and, at least according to the author, intuitive library to navigate Python source code, and code modeling based on libcst.

Enough said, I need some action!

In code as in screenwriting, it's better to show rather than tell. So, here are a couple of examples that scratch the surface of this library.

Installation

First things first, let's install the library

pip install cstq

Example base code

To start working with cstq, you can pass the path to a Python file or pass the module directly by:

In [1]: from cstq import Query
   ...: 
   ...: q = Query(
   ...:     """
   ...: import sys
   ...: 
   ...: def main() -> None:
   ...:     import os
   ...:     print('hello world' if os.environ.get("USER") else "who are you?")
   ...: 
   ...: if __name__ == "__main__":
   ...:     main()
   ...: """
   ...: )

Now lets get down to business

Basic Navigation

There are 3 types of basic navigation:

1. Direct Attribute Access

You can navigate by referencing attributes directly.

In [2]: q


Out[2]: <CollectionOfNodes nodes=['$(Module)']>

# access the body attribute of the main module
In [3]: q.body


Out[3]: <CollectionOfNodes nodes=['$(Module).body']>

# you can get all elements
In [4]: q.body[:]


Out[4]: <CollectionOfNodes nodes=['$(Module).body[0](SimpleStatementLine)', '$(Module).body[1](FunctionDef)', '$(Module).body[2](If)']>

# and get the body element of every element in the body of module 
# (if they have one)
In [5]: q.body[:].body


Out[5]: <CollectionOfNodes nodes=['$(Module).body[0](SimpleStatementLine).body', '$(Module).body[1](FunctionDef).body(IndentedBlock)', '$(Module).body[2](If).body(IndentedBlock)']>

# or just the body elements of the first element of the module
In [6]: q.body[0].body[:]


Out[6]: <CollectionOfNodes nodes=['$(Module).body[0](SimpleStatementLine).body[0](Import)']>

and then get the node (and the code) back as

In [7]: q.body[0].body[0].node()


Out[7]: Import(
    names=[
        ImportAlias(
            name=Name(
                value='sys',
                lpar=[],
                rpar=[],
            ),
            asname=None,
            comma=MaybeSentinel.DEFAULT,
        ),
    ],
    semicolon=MaybeSentinel.DEFAULT,
    whitespace_after_import=SimpleWhitespace(
        value=' ',
    ),
)

# and then get the code for that node
In [8]: q.body[0].body[0].code_for_node()


Out[8]: import sys

In [9]: q.body[0].body[:].names[0].name.node()


Out[9]: Name(
    value='sys',
    lpar=[],
    rpar=[],
)

2. Filtering

Filtering allows you to "filter" the current selection of nodes to specific ones. Each one of these accepts either a libcst.matchers or a callback (more on callbacks later). libcst.matchers presents a very powerful query language, and when that's not enough, you can always fall back to a custom callback.

In [10]: import libcst.matchers as m

# main root nodes

In [11]: q.body[:]


Out[11]: <CollectionOfNodes nodes=['$(Module).body[0](SimpleStatementLine)', '$(Module).body[1](FunctionDef)', '$(Module).body[2](If)']>

# filter out function definitions using explicit filter
In [12]: q.body[:].filter(m.FunctionDef())


Out[12]: <CollectionOfNodes nodes=['$(Module).body[1](FunctionDef)']>

# filter out function definitions using implicit filter
In [13]: q.body[m.FunctionDef()]


Out[13]: <CollectionOfNodes nodes=['$(Module).body[1](FunctionDef)']>

By using the .filter method, you can filter any selection. For instance, q.body[:] represents the elements of the body of the module. However, you can also filter out by using the __getitem__ operation ([.. add your filter here ...]), making it a bit more compact.

3. Searching

If you want to search, not only in the current selection but also at every level, you can use .search. Similar to .filter, it accepts a libcst.matchers or a callback.

In [14]: q.search(m.Import())


Out[14]: <CollectionOfNodes nodes=['$(Module).body[0](SimpleStatementLine).body[0](Import)', '$(Module).body[1](FunctionDef).body(IndentedBlock).body[0](SimpleStatementLine).body[0](Import)']>

# get the __name__ == "__main__" using search and filter
In [15]: q.search(m.If()).filter(lambda n: n.test.code() == '__name__ == "__main__"')


Out[15]: <CollectionOfNodes nodes=['$(Module).body[2](If)']>

# combining multiple search and filters into a single statement 
In [16]: q.search(m.If(), lambda n: n.test.code() == '__name__ == "__main__"')


Out[16]: <CollectionOfNodes nodes=['$(Module).body[2](If)']>

using callbacks

.filter and .search can take a callback method that takes an extended version of a CSTNode and returns true or false. The extended version of the CSTNode its the regular CSTNode with a couple of extra methods like .parent() to give your the parent, and .code() to generate the code that node represents, soo far I haven't added anything else.

codemod (changes)

If you want to change your python document, it's not hard to do, we provide a couple of simple methods like

• .change(callable, **kwargs): to change the contents of a collection of nodes.
• .replace(node): change the collection of nodes for a new one.
• .remove(): removes every node from the collection of nodes.
• .insert(index, node) : if the collection of nodes are a range, this insert the node, in the position index.
• .append(node) : add the node to the end of the range of nodes
• .extend(list[node] | collection of nodes) :extend the range to the collection or list of nodes

for instance removing the first import would be as easy as finding it (using any combination of .search and/or .filter) and then call .remove()

In [17]: q.body[0].body[0].remove()

# print the code on top
In [18]: q.code()


Out[18]: 

def main() -> None:
    import os
    print('hello world' if os.environ.get("USER") else "who are you?")

if __name__ == "__main__":
    main()

As you can see, there are several empty lines at the top. This is because of 2 things, first the module object is defined to have 1 empty line as a header, and the funtion def has the leading_lines attribute set to also have a empty line

In [19]: q.header[:].node()


Out[19]: EmptyLine(
    indent=True,
    whitespace=SimpleWhitespace(
        value='',
    ),
    comment=None,
    newline=Newline(
        value=None,
    ),
)

In [20]: q.body[0]


Out[20]: <CollectionOfNodes nodes=['$(Module).body[0](FunctionDef)']>

In [21]: q.body[0].leading_lines[:].node()


Out[21]: EmptyLine(
    indent=True,
    whitespace=SimpleWhitespace(
        value='',
    ),
    comment=None,
    newline=Newline(
        value=None,
    ),
)

Let's address this by simply changing the attribute leading_lines in that function definition using the method .change

In [22]: q.body[0].change(leading_lines=[])


Out[22]: <CollectionOfNodes nodes=['$(Module).body[0](FunctionDef)']>

In [23]: q.code()


Out[23]: 
def main() -> None:
    import os
    print('hello world' if os.environ.get("USER") else "who are you?")

if __name__ == "__main__":
    main()

for more complex changes, instead of passing the attributes to change, you can pass a callback

# reverse the order of leading lines

In [24]: q.body[0].change(lambda node: node.with_changes(leading_lines=node.leading_lines[::-1]))


Out[24]: <CollectionOfNodes nodes=['$(Module).body[0](FunctionDef)']>

you can replace a node with another one

# Lets create an "import from" node and using the serach and node function,
# and then lets use that node to replace the "import" on our module.

In [25]: import_from = Query("from python_wrapper import os").search(m.ImportFrom()).node()
    ...: 
    ...: q.search(m.Import()).replace(import_from)
    ...: q.code()


Out[25]: 
def main() -> None:
    from python_wrapper import os
    print('hello world' if os.environ.get("USER") else "who are you?")

if __name__ == "__main__":
    main()

To add a import at the top of the file, we can .insert the new node at the top, like this

# Let's add the import at the top level

In [26]: import libcst as cst
    ...: 
    ...: q.body.insert(0, cst.SimpleStatementLine(body=[import_from]))

# Let's remove the inner import
In [27]: q.search(m.FunctionDef()).search(m.ImportFrom()).remove()

# Let's print the result
In [28]: q.code()


Out[28]: 
from python_wrapper import os
def main() -> None:
    print('hello world' if os.environ.get("USER") else "who are you?")

if __name__ == "__main__":
    main()

adding a call at the end would be as easy as

# using extend to add a few lines at the end of the document


In [29]: EXTRA_LINES = Query(
    ...:     """
    ...: import my_custom_logging
    ...: my_custom_logging.log(__file__)
    ...:     """
    ...: )
    ...: 
    ...: q.body.extend(EXTRA_LINES.body[:])
    ...: 
    ...: q.code()


Out[29]: 
from python_wrapper import os
def main() -> None:
    print('hello world' if os.environ.get("USER") else "who are you?")

if __name__ == "__main__":
    main()
import my_custom_logging
my_custom_logging.log(__file__)

CST to objects and object to CST

One particular thing this library tries to do, its to give you many representations of the data, so you can interact with the source code as conformable as you can, for this sometimes you want to grab a python object from the source code and treat it as its regular object and then turn the result back to its CST representation this is not extremely hard to do

In [30]: from cstq import Query, obj2cst
    ...: import libcst.matchers as m
    ...: 
    ...: q = Query(
    ...:     """
    ...: X = [1, 2, 3, "foo", 3j]
    ...: """
    ...: )
    ...: 
    ...: the_list = q.search(m.List())
    ...: the_list


Out[30]: <CollectionOfNodes nodes=['$(Module).body[0](SimpleStatementLine).body[0](Assign).value(List)']>

# now we can turn that cst.List object into a python list using using `literal_eval_for_node`

In [31]: real_list = the_list.literal_eval_for_node()
    ...: f"type({real_list}) = {type(real_list)} "


Out[31]: type([1, 2, 3, 'foo', 3j]) = <class 'list'> 

# lets remove non integers for it and lets write it back to the cst


In [32]: only_int_list = obj2cst([e for e in real_list if isinstance(e, int)])
    ...: 
    ...: the_list.replace(only_int_list)
    ...: 
    ...: q.code()


Out[32]: 
X = [1, 2, 3]

License

cstq is distributed under the terms of the MIT license.