How to find files containing two strings in different lines

Question

A real example is: I have a Python file util.py, and I changed one of the function name from foo to bar. There might be other files using foo as function name and they are irrelevant.

I want to update all the Python files in the project from foo to bar. So I would like to do a search on files for imports such as:

import project.path.util

or

from project.path.util import foo

and any calls to foo such as one starting with:

foo(

The solution has to be whitespace tolerant. I can search each of them using grep but I am uncertain about the combined search. Any suggestions will be appreciated.

Answer 1

With GNU tools:

grep -rlZ --include='*.py' -e 'import project.path.util' \
                           -e 'from project.path.util import.*\bfoo\b' . |
  xargs -r0 sed -i 's/\bfoo\b/bar/g'

This works by asking grep to

1. search recursively (-r)
2. outputting names of matching files (-l)
3. separated by NUL instead of LF (-Z)
4. considering only files ending in '.py' (--include='*.py')

and asking sed to perform the replacement in-place (-i) in all matching files, if there are any (xargs -r).

Answer 2

This should do.

/g will substitute all instances of foo with bar

find <Project root dir> -type f -exec sed -i "{}" -e 's/foo/bar/g' \;

If you want to restrict the change to *.py files only :

find <Project root dir> -type f -name "*.py" -exec sed -i "{}" -e 's/foo/bar/g' \;

If there are instances of foo that should not be replaced, you can restrict to 4 cases:

• import foo
• from module import foo
• def foo
• foo(

But not

• damned_foo(
• damnedfoo(
• damned1foo(
• damned-foo(

Then the command would look like this :

foo="foo";bar="bar";find <Project root dir>  -type f -exec sed -i "{}" \
-r -e "s/[^a-zA-Z0-9_\-]$foo\(/$bar\(/g" -e "s/def $foo/def $bar/g" \
-e "s/import $foo/import $bar/g" \;

Answer 3

Suppose you are in the directory that you want to search from (it matches any subdirectory also):

 for i in `find . -type f -name '*.py' -exec grep -lr 'foo' {} +`; do sed -i 's/foo/bar/' $i; done

Edit:

In order to avoid matching functions containing the characters, like "damned_fools" we can improve it with \b to match word boundaries:

for i in `find . -type f -name '*.py' -exec grep -lr 'foo' {} +`; do sed -i 's/\bfoo\b/bar/' $i; done

Reference: https://stackoverflow.com/questions/1032023/sed-whole-word-search-and-replace