使用sed在条件下解释多行

Question

I'm stuck on constructing a sed expression that will parse a python file's imports and extract the names of the modules.

This is a simple example that I solved using (I need the output to be the module names without 'as' or any spaces..):

from testfunctions import mod1, mod2 as blala, mod3, mod4

What I have so far:

grep -ir "from testfunctions import" */*.py | sed -E s/'\s+as\s+\w+'//g | sed -E s/'from testfunctions import\s+'//g

This does get me the required result in a situation as above.

The problem: In files where the imports are like so:

from testfunctions import mod1, mod2 as blala, mod3, mod4 \
     mod5, mod6 as bla, mod7 \
   mod8, mod9 ...

Any ideas how I can improve my piped expression to handle multiple lines?

Answer 1

尝试这个;

   sed -n -r '/from/,/^\s*$/p;' *.py | sed ':x; /\\$/ { N; s/\\\n//; tx }'  | sed 's/^.*.import//g;s/  */ /g'

Answer 2

Thanks everyone for your help. I didn't know a module such as ast exists.. It really helped me achieve my goal.

I put together a simple version of the solution I needed, just for reference if anyone else encounters this question as well:

import glob
import ast

moduleList = []
# get all .py file names
testFiles = glob.glob('*/*.py')
for testFile in testFiles:
    with open(testFile) as code:
        # ast.parse creates the tree off of plain code
        tree = ast.parse(code.read())
        # there are better ways to traverse the tree, in this sample there
        # is no guarantee to the traversal order
        for node in ast.walk(tree):
            if isinstance(node, ast.ImportFrom) and node.module == 'testfunctions':
                # each node will contain an ast.ImportFrom instance which
                # data members are: module, names(list of ast.alias) and level
                moduleList.extend([alias.name for alias in node.names])

You can read more about it in (probably the only detailed page about ast in the whole web) here: https://greentreesnakes.readthedocs.io/en/latest/manipulating.html#inspecting-nodes