Python多行正则表达式+多项一次性读取文件

Question

//Last modified: Sat, Apr 16, 2011 09:55:04 AM
//Codeset: ISO-8859-1
fileInfo "version" "20x64";
createNode newnode -n "a_SET";
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" "blabla";
    setAttr -l on -k on ".test2" -type "string" "blablabla";
createNode newnode -n "b_SET";
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" "hmm";
    setAttr -l on -k on ".test2" -type "string" "ehmehm";

in Python:

I need to read the newnode names for instance "a_SET" and "b_SET" and their corresponding attribute values so {"a_SET": {"test1":"blabla", "test2":"blablabla"} and the same for the b_SET - there could be unknown amount of sets - like c_SET d_SET etc.

I've tried looping through lines and matching it there:

for line in fileopened:
    setmatch = re.match( r'^(createNode set -n ")(.*)(_SET)(.*)' , line)
     if setmatch:
            sets.append(setmatch.group(2))

and as soon as I find a match here I would loop through next lines to get the attributes (test1, test2) for that set until I find a new set - for instance c_SET or an EOF.

What would be the best way to grab all that info in one go with the re.MULTILINE?

Answer 1

You can use regexp positive lookahead to split the groups:

(yourGroupSeparator)(.*?)(?=yourGroupSeparator|\Z)

In your example:

import re

lines = open("e:/temp/test.txt").read()
matches = re.findall(r'createNode newnode \-n (\"._SET\");(.*?)(?=createNode|\Z)', lines, re.MULTILINE + re.DOTALL);

for m in matches:
    print "%s:" % m[0], m[1]


"""
Result:
>>>
"a_SET":
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" "blabla";
    setAttr -l on -k on ".test2" -type "string" "blablabla";

"b_SET":
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" "hmm";
    setAttr -l on -k on ".test2" -type "string" "ehmehm";
"""

If you want the results on a dict, you can use:

result = {}
for k, v in matches:
    result[k] = v   # or maybe v.split() or v.split(";")

after findall

Answer 2

I got this:

import re

filename = 'tr.txt'

with open(filename,'r') as f:
    ch = f.read()

pat = re.compile('createNode newnode -n ("\w+?_SET");(.*?)(?=createNode|\Z)',re.DOTALL)
pit = re.compile('^ *setAttr.+?("[^"\n]+").+("[^"\n]+");(?:\n|\Z)',re.MULTILINE)

dic = dict( (mat.group(1),dict(pit.findall(mat.group(2)))) for mat in pat.finditer(ch)) 
print dic

result

{'"b_SET"': {'".test2"': '"ehmehm"', '".test1"': '"hmm"'}, '"a_SET"': {'".test2"': '"blablabla"', '".test1"': '"blabla"'}}

.

Question:

what if there must be character '"' in the strings ? How is it represented ?

.

EDIT

I had some difficulty to find the solution because I didn't choose the facility.

Here's a new pattern that catches the FIRST string "..." and the LAST string "..." present after a string " setAttr" and before the next " setAttr" . So several "..." can be present , not only 3. You didn't asked this condition, but I thought it may happen to be needed.

I also managed to make possible the presence of newlines in the strings to catch "....\\n......" , not only around them. For that , I was obliged to invent something new for me: (?:\\n(?! *setAttr)|[^"\\n]) that means : all characters, except '"' and common newlines \\n , are accepted and also only the newlines that are not followed by a line beginning with ' *setAttr'

For (?:\\n(?! *setAttr)|.) it means : newlines not followed by a line beginning with ' *setAttr' and all the other non-newline characters.

Hence, any other special sequence as tab or whatever else are automatically accpted in the matchings.

ch = '''//Last modified: Sat, Apr 16, 2011 09:55:04 AM
//Codeset: ISO-8859-1
fileInfo "version" "20x64";
createNode newnode -n "a_SET";
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" "blabla";
    setAttr -l on -k on ".test2" -type "string" "blablabla";
createNode newnode -n "b_SET";
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" (
      "hmm bl
      abla\tbla" );
    setAttr -l on -k on ".tes\nt\t2" -type "string" "ehm\tehm";
    setAttr -l on -k on ".test3" -type "string" "too
    much" "pff" """ "feretini" "gol\nolo";
    '''

import re

pat = re.compile('createNode newnode -n ("\w+?_SET");(.*?)(?=createNode|\Z)',re.DOTALL)
pot = re.compile('^ *setAttr.+?'
                 '"((?:\n(?! *setAttr)|[^"\n])+)"'
                 '(?:\n(?! *setAttr)|.)+'
                 '"((?:\n(?! *setAttr)|[^"\n])+)"'
                 '.*;(?:\n|\Z)',re.MULTILINE)

dic = dict( (mat.group(1),dict(pot.findall(mat.group(2)))) for mat in pat.finditer(ch)) 
for x in dic:
    print x,'\n',dic[x],'\n'

result

"b_SET" 
{'.test3': 'gol\nolo', '.test1': 'hmm bl\n      abla\tbla', '.tes\nt\t2': 'ehm\tehm'} 

"a_SET" 
{'.test1': 'blabla', '.test2': 'blablabla'}

Answer 3

Another possible option:

createNode newnode -n "b_SET";
    addAttr -ci true -k true -sn "connections" -ln "connections" -dt "string";
    setAttr -l on -k off ".tx";
    setAttr -l on -k off ".ty";
    setAttr -l on -k off ".sz";
    setAttr -l on -k on ".test1" -type "string" (
      "hmm blablabla" );
    setAttr -l on -k on ".test2" -type "string" "ehmehm";

So as you can see ".test1" value is now split with a /n line separator. How would you go around that using eyquem's approach?

pit = re.compile('^ *setAttr.+?("[^"\n]+").+("[^"\n]+");(?:\n|\Z)',re.MULTILINE)