REGEX in Python: what's wrong with (?<!\\)\".+(?<!\\)\"? - python

trying to parse JSON key names within quotes, including escaped quotes.
my thinking is: take anything between quotes not prefixed with \
(?<!\\)\".+(?<!\\)\"
where (?<!\\)\" should screen for " but not \" but Python complains about unbalanced parenthesis.
if I use (?<!\\\)\" Python is happy , but this doesn't work:
re.findall('(?<!\\\)\".+(?<!\\\)\"','"this is \"the\". key"."and this.is.the.child"')
leads:
['"this is "the". key"."and this.is.the.child"']
when I expect:
['"this is "the". key"', '"and this.is.the.child"']
split at the dot which is enclosed with " without escape.
I feel like i need an 'anything but not escaped double quote ' in the middle, but if
[^"] screens for anything but a double quote, I don't know how to negate the (?<!\\\)\" expression within a [ ] set that takes characters as literals.
i would want something like [^(?<!\\\)\"] but that doesn't work.
I tried things like [[^"]|(\")]+ (anything but a double quote, or a \" ) but that doesn't seem to work either...
Can;t seem to find the right way to do this...
Any ideas?
Thanks for help
EDIT:
My real goal is to be able to split full 'text' JSON key names to transform them into alphanum only values. The transform is irrelevant here, but the goal is to split the keys to represent the hierarchy properly. The keys are in text form.
EDIT 2:
even though OmnipotentEntity is most likely right, writing a parser will have to wait..
This solution below doesn't support the "\" or "\\" cases as indicated in his comments.
I settled with
"(?:\\"|[^"])*?"|(?<=\.)[^".]+?(?=\.)|^[^".]+?(?=\.)|(?<=\.)[^".]+?$
inspired by the answer from Avinash Raj
but adding support for keys that are not enclosed in double quotes:
no quotes beginning of line ending with .
.key.
and
.lastkey
when substituting [empty] with the same regex, one should find 1 less element than the number of found strings, or there is an error.
something like .. outside "" will fail that test

Fundamentally, using a regular expression to match quoted strings is impossible in the general case. JSON is not a regular language (all regular languages are LL(1) but not all LL(1) languages are regular, JSON is one of these), so it cannot be matched by a regular expression.
Avinash Raj's regular expression (?<!\\)".*?(?<!\\)", for instance, fails on the the case "\\". Because the quote is preceded by a \ but the backslash doesn't function as an escape. But you can't special case this situation because then "\\\"" will fail. And if you special case this situation, you can just use 4 \ and then 5 \ etc.
Lookbehinds aren't part of standard regular expressions so they can match more grammars than simply regular ones. So you might be able to come up with a regular expression that works in this case. However, I would recommend writing a parser instead, they are very easy to do for LL(1) grammars. It will be easier, more understandable, less brittle, and give you more leverage to deal with non-conformant JSON and give you the ability to write better diagnostic messages in this case.

Try to define your regex as raw string notation.
>>> s = r'"this is \"the\". key"."and this.is.the.child"'
>>> re.findall(r'"(?:\\"|[^"])*?"', s)
['"this is \\"the\\". key"', '"and this.is.the.child"']
DEMO
OR
>>> re.findall(r'(?<!\\)".*?(?<!\\)"', s)
['"this is \\"the\\". key"', '"and this.is.the.child"']
(?<!\\) called negative lookbehind which asserts that the match won't be preceded by a backslash.
" Matches a double quotes.
.*?(?<!\\)" Matches all the characters non-greedily upto the double quotes which is not preceded by a backslash.

Related

How to write a regular expression with a list of quotes, using raw string literal? [duplicate]

Technically, any odd number of backslashes, as described in the documentation.
>>> r'\'
File "<stdin>", line 1
r'\'
^
SyntaxError: EOL while scanning string literal
>>> r'\\'
'\\\\'
>>> r'\\\'
File "<stdin>", line 1
r'\\\'
^
SyntaxError: EOL while scanning string literal
It seems like the parser could just treat backslashes in raw strings as regular characters (isn't that what raw strings are all about?), but I'm probably missing something obvious.
The whole misconception about python's raw strings is that most of people think that backslash (within a raw string) is just a regular character as all others. It is NOT. The key to understand is this python's tutorial sequence:
When an 'r' or 'R' prefix is present, a character following a
backslash is included in the string without change, and all
backslashes are left in the string
So any character following a backslash is part of raw string. Once parser enters a raw string (non Unicode one) and encounters a backslash it knows there are 2 characters (a backslash and a char following it).
This way:
r'abc\d' comprises a, b, c, \, d
r'abc\'d' comprises a, b, c, \, ', d
r'abc\'' comprises a, b, c, \, '
and:
r'abc\' comprises a, b, c, \, ' but there is no terminating quote now.
Last case shows that according to documentation now a parser cannot find closing quote as the last quote you see above is part of the string i.e. backslash cannot be last here as it will 'devour' string closing char.
The reason is explained in the part of that section which I highlighted in bold:
String quotes can be escaped with a
backslash, but the backslash remains
in the string; for example, r"\"" is a
valid string literal consisting of two
characters: a backslash and a double
quote; r"\" is not a valid string
literal (even a raw string cannot end
in an odd number of backslashes).
Specifically, a raw string cannot end
in a single backslash (since the
backslash would escape the following
quote character). Note also that a
single backslash followed by a newline
is interpreted as those two characters
as part of the string, not as a line
continuation.
So raw strings are not 100% raw, there is still some rudimentary backslash-processing.
That's the way it is! I see it as one of those small defects in python!
I don't think there's a good reason for it, but it's definitely not parsing; it's really easy to parse raw strings with \ as a last character.
The catch is, if you allow \ to be the last character in a raw string then you won't be able to put " inside a raw string. It seems python went with allowing " instead of allowing \ as the last character.
However, this shouldn't cause any trouble.
If you're worried about not being able to easily write windows folder pathes such as c:\mypath\ then worry not, for, you can represent them as r"C:\mypath", and, if you need to append a subdirectory name, don't do it with string concatenation, for it's not the right way to do it anyway! use os.path.join
>>> import os
>>> os.path.join(r"C:\mypath", "subfolder")
'C:\\mypath\\subfolder'
In order for you to end a raw string with a slash I suggest you can use this trick:
>>> print r"c:\test"'\\'
test\
It uses the implicit concatenation of string literals in Python and concatenates one string delimited with double quotes with another that is delimited by single quotes. Ugly, but works.
Another trick is to use chr(92) as it evaluates to "\".
I recently had to clean a string of backslashes and the following did the trick:
CleanString = DirtyString.replace(chr(92),'')
I realize that this does not take care of the "why" but the thread attracts many people looking for a solution to an immediate problem.
Since \" is allowed inside the raw string. Then it can't be used to identify the end of the string literal.
Why not stop parsing the string literal when you encounter the first "?
If that was the case, then \" wouldn't be allowed inside the string literal. But it is.
The reason for why r'\' is syntactical incorrect is that although the string expression is raw the used quotes (single or double) always have to be escape since they would mark the end of the quote otherwise. So if you want to express a single quote inside single quoted string, there is no other way than using \'. Same applies for double quotes.
But you could use:
'\\'
Another user who has since deleted their answer (not sure if they'd like to be credited) suggested that the Python language designers may be able to simplify the parser design by using the same parsing rules and expanding escaped characters to raw form as an afterthought (if the literal was marked as raw).
I thought it was an interesting idea and am including it as community wiki for posterity.
Naive raw strings
The naive idea of a raw string is
If I put an r in front of a pair of quotes,
I can put whatever I want between the quotes
and it will mean itself.
Unfortunately, this does not work, because if the whatever
happens to contain a quote, the raw string would end at that point.
It is simply impossible that I can put "whatever I want"
between fixed delimiters, because some of it could look like
the terminating delimiter -- no matter what that delimiter is.
Real-world raw strings (variant 1)
One possible approach to this problem would be to say
If I put an r in front of a pair of quotes,
I can put whatever I want between the quotes
as long as it does not contain a quote
and it will mean itself.
This restriction sounds harsh, until one recognizes that
Python's large offering of quotes can accommodate most situations
with this rule. The following are all valid Python quotes:
'
"
'''
"""
With this many possibilities for the delimiter, almost anything
can be made to work.
About the only exception would be if the string
literal is supposed to contain a complete list of all allowed
Python quotes.
Real-world raw strings (variant 2, as in Python)
Python, however, takes a different route using
an extended version of the above rule.
It effectively states
If I put an r in front of a pair of quotes,
I can put whatever I want between the quotes
as long as it does not contain a quote
and it will mean itself.
If I insist on including a quote, even that is allowed,
but I have to put a backslash before it.
So the Python approach is, in a sense, even more liberal
than variant 1 above -- but it has the side effect of
"mis"interpreting the closing quote as part of the string
if the last intended character of the string is a backslash.
Variant 2 is not helpful:
If I want the quote in my string,
but not the backslash, the allowed version of my string literal
will not be what I need.
However, given the three different other kinds of quotes I have
at my disposal, I will probably just pick one of those and my
problem will be solved -- so this is not problematic case.
The problematic case is this one:
If I want my string to end with a backslash, I am at a loss.
I need to resort to concatenating a non-raw string literal
containing the backslash.
Conclusion
After writing this, I go with several of the other posters
that variant 1 would have been easier to understand and to accept
and therefore more pythonic. That's life!
Comming from C it pretty clear to me that a single \ works as escape character allowing you to put special characters such as newlines, tabs and quotes into strings.
That does indeed disallow \ as last character since it will escape the " and make the parser choke. But as pointed out earlier \ is legal.
some tips :
1) if you need to manipulate backslash for path then standard python module os.path is your friend. for example :
os.path.normpath('c:/folder1/')
2) if you want to build strings with backslash in it BUT without backslash at the END of your string then raw string is your friend (use 'r' prefix before your literal string). for example :
r'\one \two \three'
3) if you need to prefix a string in a variable X with a backslash then you can do this :
X='dummy'
bs=r'\ ' # don't forget the space after backslash or you will get EOL error
X2=bs[0]+X # X2 now contains \dummy
4) if you need to create a string with a backslash at the end then combine tip 2 and 3 :
voice_name='upper'
lilypond_display=r'\DisplayLilyMusic \ ' # don't forget the space at the end
lilypond_statement=lilypond_display[:-1]+voice_name
now lilypond_statement contains "\DisplayLilyMusic \upper"
long live python ! :)
n3on
Despite its role, even a raw string cannot end in a single
backslash, because the backslash escapes the following quote
character—you still must escape the surrounding quote character to
embed it in the string. That is, r"...\" is not a valid string
literal—a raw string cannot end in an odd number of backslashes.
If you need to end a raw string with a single backslash, you can use
two and slice off the second.
I encountered this problem and found a partial solution which is good for some cases. Despite python not being able to end a string with a single backslash, it can be serialized and saved in a text file with a single backslash at the end. Therefore if what you need is saving a text with a single backslash on you computer, it is possible:
x = 'a string\\'
x
'a string\\'
# Now save it in a text file and it will appear with a single backslash:
with open("my_file.txt", 'w') as h:
h.write(x)
BTW it is not working with json if you dump it using python's json library.
Finally, I work with Spyder, and I noticed that if I open the variable in spider's text editor by double clicking on its name in the variable explorer, it is presented with a single backslash and can be copied to the clipboard that way (it's not very helpful for most needs but maybe for some..).

Regex whitespace, brackets, and parens

I am trying to match a string in the following form:
require([ "foo/bar", "foo2/bar2" ])
Whitespace should be ignored entirely. I am using the following regex with little success:
require\\(\s*\\[[.\s]*\\]\\)
Any suggestions? I know that regex attempt is horrible...
EDIT: I am using Python!
If you are using Java or PHP with double-quoted strings or somethig similar, you need to double escape the \s as well. If not, then you need to remove all double backslashes instead (and make them single backslashes). Also note, that [.\s] matches only periods and whitespace (. loses its wildcard meaning within character classes). If you really want to match anything use [\s\S] instead.
Assuming double escaping is required in the language you use:
require\\(\\s*\\[[\\S\\s]*\\]\\)
Note that this will cause problems if this occurs multiple times in the same string. Then you would get a match from the first require([ to the last ]). To avoid this, disallow ] within the repetition. However, be aware that this in turn can cause problems if your strings within require may contain ] themselves:
require\\(\\s*\\[[^]]*\\]\\)

python regex re.compile match

I am trying to match (using regex in python):
http://images.mymaterials.com/images/steel-images/small/steel/steel800/steel800-2.jpg
in the following string:
http://www.mymaterialssite.com','http://images.mymaterials.com/images/steel-images/small/steel/steel800/steel800-2.jpg','Model Photo'
My code has something like this:
temp="http://www.mymaterialssite.com','http://images.mymaterials.com/images/steel-images/small/steel/steel800/steel800-2.jpg','Model Photo'"
dummy=str(re.compile(r'.com'',,''(.*?)'',,''Model Photo').search(str(temp)).group(1))
I do not think the "dummy" is correct & I am unsure how I "escape" the single and double quotes in the regex re.compile command.
I tried googling for the problem, but I couldnt find anything relevant.
Would appreciate any guidance on this.
Thanks.
The easiest way to deal with strings in Python that contain escape characters and quotes is to triple double-quote the string (""") and prefix it with r. For example:
my_str = r"""This string would "really "suck"" to write if I didn't
know how to tell Python to parse it as "raw" text with the 'r' character and
triple " quotes. Especially since I want \n to show up as a backlash followed
by n. I don't want \0 to be the null byte either!"""
The r means "take escape characters as literal". The triple double-quotes (""") prevent single-quotes, double-quotes, and double double-quotes from prematurely ending the string.
EDIT: I expanded the example to include things like \0 and \n. In a normal string (not a raw string) a \ (the escape character) signifies that the next character has special meaning. For example \n means "the newline character". If you literally wanted the character \ followed by n in your string you would have to write \\n, or just use a raw string instead, as I show in the example above.
You can also read about string literals in the Python documentation here:
For beginners: http://docs.python.org/tutorial/introduction.html#strings
Complex explanation: http://docs.python.org/reference/lexical_analysis.html#string-literals
Try triple quotes:
import re
tmp=""".*http://images.mymaterials.com/images/steel-images/small/steel/steel800/steel800-2.jpg.*"""
str="""http://www.mymaterialssite.com\'\,\'http://images.mymaterials.com/images/steel-images/small/steel/steel800/steel800-2.jpg','Model Photo'"""
x=re.match(tmp,str)
if x!=None:
print x.group()
Also you were missing the .* in the beginning of the pattern and at the end. I added that too.
if you use double quotes (which have the same meaning as the single ones, in Python), you don't have to escape at all.. (in this case). you can even use string literal without the starting r (you don't have any backslash there)
re.compile(".com','(.*?)','Model Photo")
Commas don't need to be escaped, and single quotes don't need to be escaped if you use double quotes to create the string:
>>> dummy=re.compile(r".com','(.*?)','Model Photo").search(temp).group(1)
>>> print dummy
http://images.mymaterials.com/images/steel-images/small/steel/steel800/steel800-2.jpg
Note that I also removed some unnecessary str() calls, and for future reference if you do ever need to escape single or double quotes (say your string contains both), use a backslash like this:
'.com\',\'(.*?)\',\'Model Photo'
As mykhal pointed out in comments, this doesn't work very nicely with regex because you can no longer use the raw string (r'...') literal. A better solution would be to use triple quoted strings as other answers suggested.

python "re" package, strange phenomenon with "raw" string

I am seeing the following phenomenon, couldn't seem to figure it out, and didn't find anything with some search through archives:
if I type in:
>>> if re.search(r'\n',r'this\nis\nit'):<br>
... print 'found it!'<br>
... else:<br>
... print "didn't find it"<br>
...
I will get:
didn't find it!
However, if I type in:
>>> if re.search(r'\\n',r'this\nis\nit'):<br>
... print 'found it!'<br>
... else:<br>
... print "didn't find it"<br>
...
Then I will get:
found it!
(The first one only has one backslash on the r'\n' whereas the second one has two backslashes in a row on the r'\\n' ... even this interpreter is removing one of them.)
I can guess what is going on, but I don't understand the official mechanism as to why this is happening: in the first case, I need to escape two things: both the regular expression and the special strings. "Raw" lets me escape the special strings, but not the regular expression.
But there will never be a regular expression in the second string, since it is the string being matched. So there is only a need to escape once.
However, something doesn't seem consistent to me: how am I supposed to ensure that the characters REALLY ARE taken literally in the first case? Can I type rr'' ? Or do I have to ensure that I escape things twice?
On a similar vein, how do I ensure that a variable is taken literally (or that it is NOT taken literally)? E.g., what if I had a variable tmp = 'this\nis\nmy\nhome', and I really wanted to find the literal combination of a slash and an 'n', instead of a newline?
Thanks!Mike
re.search(r'\n', r'this\nis\nit')
As you said, "there will never be a regular expression in the second string." So we need to look at these strings differently: the first string is a regex, the second just a string. Usually your second string will not be raw, so any backslashes are Python-escapes, not regex-escapes.
So the first string consists of a literal "\" and an "n". This is interpreted by the regex parser as a newline (docs: "Most of the standard escapes supported by Python string literals are also accepted by the regular expression parser"). So your regex will be searching for a newline character.
Your second string consists of the string "this" followed by a literal "\" and an "n". So this string does not contain an actual newline character. Your regex will not match.
As for your second regex:
re.search(r'\\n', r'this\nis\nit')
This version matches because your regex contains three characters: a literal "\", another literal "\" and an "n". The regex parser interprets the two slashes as a single "\" character, followed by an "n". So your regex will be searching for a "\" followed by an "n", which is found within the string. But that isn't very helpful, since it has nothing to do with newlines.
Most likely what you want is to drop the r from the second string, thus treating it as a normal Python string.
re.search(r'\n', 'this\nis\nit')
In this case, your regex (as before) is searching for a newline character. And, it finds it, because the second string contains the word "this" followed by a newline.
Escaping special sequences in string literals is one thing, escaping regular expression special characters is another. The row string modifier only effects the former.
Technically, re.search accepts two strings and passes the first to the regex builder with re.compile. The compiled regex object is used to search patterns inside simple strings. The second string is never compiled and thus it is not subject to regex special character rules.
If the regex builder receives a \n after the string literal is processed, it converts this sequence to a newline character. You also have to escape it if you need the match the sequence instead.
All rationale behind this is that regular expressions are not part of the language syntax. They are rather handled within the standard library inside the re module with common building blocks of the language.
The re.compile function uses special characters and escaping rules compatible with most commonly used regex implementations. However, the Python interpreter is not aware of the whole regular expression concept and it does not know whether a string literal will be compiled into a regex object or not. As a result, Python can't provide any kind syntax simplification such as the ones you suggested.
Regexes have their own meaning for literal backslashes, as character classes like \d. If you actually want a literal backslash character, you will in fact need to double-escape it. It's really not supposed to be parallel since you're comparing a regex to a string.
Raw strings are just a convenience, and it would be way overkill to have double-raw strings.

How to write a regular expression to match a string literal where the escape is a doubling of the quote character?

I am writing a parser using ply that needs to identify FORTRAN string literals. These are quoted with single quotes with the escape character being doubled single quotes. i.e.
'I don''t understand what you mean'
is a valid escaped FORTRAN string.
Ply takes input in regular expression. My attempt so far does not work and I don't understand why.
t_STRING_LITERAL = r"'[^('')]*'"
Any ideas?
A string literal is:
An open single-quote, followed by:
Any number of doubled-single-quotes and non-single-quotes, then
A close single quote.
Thus, our regex is:
r"'(''|[^'])*'"
You want something like this:
r"'([^']|'')*'"
This says that inside of the single quotes you can have either double quotes or a non-quote character.
The brackets define a character class, in which you list the characters that may or may not match. It doesn't allow anything more complicated than that, so trying to use parentheses and match a multiple-character sequence ('') doesn't work. Instead your [^('')] character class is equivalent to [^'()], i.e. it matches anything that's not a single quote or a left or right parenthesis.
It's usually easy to get something quick-and-dirty for parsing particular string literals that are giving you problems, but for a general solution you can get a very powerful and complete regex for string literals from the pyparsing module:
>>> import pyparsing
>>> pyparsing.quotedString.reString
'(?:"(?:[^"\\n\\r\\\\]|(?:"")|(?:\\\\x[0-9a-fA-F]+)|(?:\\\\.))*")|(?:\'(?:[^\'\\n\\r\\\\]|(?:\'\')|(?:\\\\x[0-9a-fA-F]+)|(?:\\\\.))*\')'
I'm not sure about significant differences between FORTRAN's string literals and Python's, but it's a handy reference if nothing else.
import re
ch ="'I don''t understand what you mean' and you' ?"
print re.search("'.*?'",ch).group()
print re.search("'.*?(?<!')'(?!')",ch).group()
result
'I don'
'I don''t understand what you mean'

Categories