472,378 Members | 1,498 Online

I'm trying to develop a little script that does some string
manipulation. I have some few hundred strings that currently look like
this:

cond(a,b,c)

and I want them to look like this:

cond(c,a,b)

but it gets a little more complicated because the conds themselves may
have conds within, like the following:

cond(0,cond(c,cond(e,cond(g,h,(a<f)),(a<d)),(a<b)) ,(a<1))

What I want to do in this case is move the last parameter to the front
and then work backwards all the way out (if you're thinking recursion
too, I'm vindicated) so that it ends up looking like this:

cond((a<1), 0, cond((a<b),c,cond((a<d), e, cond((a<f), g, h))))

futhermore, the conds may be multiplied by an expression, such as the
following:

cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float(a))

Here, all I want to do is switch the parameters of the conds without
touching the expression, like so:

cond(f,-1,1)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4)+ (float(b)*2)+float(a))

So that's the gist of my problem statement. I immediately thought that
regular expressions would provide an elegant solution. I would go
through the string by conds, stripping them & the () off, until I got
to the lowest level, then move the parameters and work backwards. That
thought process became this:
-------------------------------------CODE--------------------------------------------------------
import re

def swap(left, middle, right):
left = left.replace("(", "")
right = right.replace(")", "")
temp = left
left = right
right = temp
temp = middle
middle = right
right = temp
whole = 'cond(' + left + ',' + middle + ',' + right + ')'
return whole

def condReplacer(string):
#regex = re.compile(r'cond\(.*,.*,.+\)')
regex = re.compile(r'cond\(.*,.*,.+?\)')
if not regex.search(string):
print "whole string is: " + string
[left, middle, right] = string.split(',')
right = right.replace('\'', ' ')
string = swap(left.strip(), middle.strip(), right.strip())
print "the new string is:" + string
return string
else:
more_conds = regex.search(string)
temp_string = more_conds.group()
firstParen = temp_string.find('(')
temp_string = temp_string[firstParen:]
print "there are more conditionals!" + temp_string
condReplacer(temp_string)
for line in file:
regex = r'cond\(.*,.*,.+\)?'
if re.search(regex,line,re.DOTALL):
condReplacer(line)

if __name__ == "__main__":
input_file = open("only_conds2.txt", 'r')
-------------------------------------CODE--------------------------------------------------------

I think my problem lies in my regular expression... If I use the one
commented out I do a greedy search and in my test case where I have a
conditional * an expression, I grab the expression too, like so:

INPUT:

cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float(a))
OUTPUT:
whole string is:
(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float
(a))
the new string
is:cond(f*((float(e*(2**4+(float(d*8+(float(c*4+(f loat(b*2+float
(a,-1,1)

when all I really want to do is grab the part associated with the cond.
But if I do a non-greedy search I avoid that problem but stop too early
when I have an expression like this:

INPUT:
cond(a,b,(abs(c) >= d))
OUTPUT:
whole string is: (a,b,(abs(c)
the new string is:cond((abs(c,a,b)

Can anyone help me with the regular expression? Is this even the best
approach to take? Anyone have any thoughts?

Aug 23 '06 #1
4 1816
cond(a,b,c)
>
and I want them to look like this:

cond(c,a,b)

but it gets a little more complicated because the conds themselves may
have conds within, like the following:

cond(0,cond(c,cond(e,cond(g,h,(a<f)),(a<d)),(a<b)) ,(a<1))
Regexps are *really* *REALLY* *bad* at arbitrarily nested
structures. really.

Sounds more like you want something like a lex/yacc sort of
solution. IIUC, pyparsing may do the trick for you. I'm not a
pyparsing wonk, but I can hold my own when it comes to crazy
regexps, and can tell you from experience that regexps are *not*
a good path to try and go down for this problem.

Many times, a regexp can be hammered into solving problems
superior solutions than employing regexps. This case is not even
one of those.

If you know the maximum depth of nesting you'll encounter, you
can do some hackish stunts to shoehorn regexps to solve the
problem. But if they are truely of arbitrary nesting-depth,
*good* *luck*! :)

-tkc

Aug 23 '06 #2
MooMaster wrote:
I'm trying to develop a little script that does some string
manipulation. I have some few hundred strings that currently look like
this:

cond(a,b,c)

and I want them to look like this:

cond(c,a,b)

but it gets a little more complicated because the conds themselves may
have conds within, like the following:

cond(0,cond(c,cond(e,cond(g,h,(a<f)),(a<d)),(a<b)) ,(a<1))

What I want to do in this case is move the last parameter to the front
and then work backwards all the way out (if you're thinking recursion
too, I'm vindicated) so that it ends up looking like this:

cond((a<1), 0, cond((a<b),c,cond((a<d), e, cond((a<f), g, h))))

futhermore, the conds may be multiplied by an expression, such as the
following:

cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float(a))

Here, all I want to do is switch the parameters of the conds without
touching the expression, like so:

cond(f,-1,1)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4)+ (float(b)*2)+float(a))

So that's the gist of my problem statement. I immediately thought that
regular expressions would provide an elegant solution. I would go
through the string by conds, stripping them & the () off, until I got
to the lowest level, then move the parameters and work backwards. That
thought process became this:
-------------------------------------CODE--------------------------------------------------------
import re

def swap(left, middle, right):
left = left.replace("(", "")
right = right.replace(")", "")
temp = left
left = right
right = temp
temp = middle
middle = right
right = temp
whole = 'cond(' + left + ',' + middle + ',' + right + ')'
return whole

def condReplacer(string):
#regex = re.compile(r'cond\(.*,.*,.+\)')
regex = re.compile(r'cond\(.*,.*,.+?\)')
if not regex.search(string):
print "whole string is: " + string
[left, middle, right] = string.split(',')
right = right.replace('\'', ' ')
string = swap(left.strip(), middle.strip(), right.strip())
print "the new string is:" + string
return string
else:
more_conds = regex.search(string)
temp_string = more_conds.group()
firstParen = temp_string.find('(')
temp_string = temp_string[firstParen:]
print "there are more conditionals!" + temp_string
condReplacer(temp_string)
for line in file:
regex = r'cond\(.*,.*,.+\)?'
if re.search(regex,line,re.DOTALL):
condReplacer(line)

if __name__ == "__main__":
input_file = open("only_conds2.txt", 'r')
-------------------------------------CODE--------------------------------------------------------

I think my problem lies in my regular expression... If I use the one
commented out I do a greedy search and in my test case where I have a
conditional * an expression, I grab the expression too, like so:

INPUT:

cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float(a))
OUTPUT:
whole string is:
(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float
(a))
the new string
is:cond(f*((float(e*(2**4+(float(d*8+(float(c*4+(f loat(b*2+float
(a,-1,1)

when all I really want to do is grab the part associated with the cond.
But if I do a non-greedy search I avoid that problem but stop too early
when I have an expression like this:

INPUT:
cond(a,b,(abs(c) >= d))
OUTPUT:
whole string is: (a,b,(abs(c)
the new string is:cond((abs(c,a,b)

Can anyone help me with the regular expression? Is this even the best
approach to take? Anyone have any thoughts?

You're gonna want a parser for this. pyparsing or spark would suffice.
However, since it looks like your source strings are valid python you
could get some traction out of the tokenize standard library module:

from tokenize import generate_tokens
from StringIO import StringIO

s =
'cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float(a))'

print t[1],
Prints:
cond ( - 1 , 1 , f ) * ( ( float ( e ) * ( 2 ** 4 ) ) + ( float ( d ) *
8 ) + ( float ( c ) * 4 ) + ( float ( b ) * 2 ) + float ( a ) )

Once you've got that far the rest should be easy. :)

Peace,
~Simon

http://pyparsing.wikispaces.com/
http://pages.cpsc.ucalgary.ca/~aycock/spark/
http://docs.python.org/lib/module-tokenize.html

Aug 23 '06 #3
"MooMaster" <nt*****@gmail.comwrote in message
I'm trying to develop a little script that does some string
manipulation. I have some few hundred strings that currently look like
this:

cond(a,b,c)

and I want them to look like this:

cond(c,a,b)
<snip>

Pyparsing makes this a fairly tractable problem. The hardest part is
defining the valid contents of a relational and arithmetic expression, which
may be found within the arguments of your cond(a,b,c) constructs.

Not guaranteeing this 100%, but it did convert your pathologically nested
example on the first try.

-- Paul

----------
from pyparsing import *

ident = ~Literal("cond") + Word(alphas)
number = Combine(Optional("-") + Word(nums) + Optional("." + Word(nums)))

arithExpr = Forward()
funcCall = ident+"("+delimitedList(arithExpr)+")"
operand = number | funcCall | ident
binop = oneOf("+ - * /")
arithExpr << ( ( operand + ZeroOrMore( binop + operand ) ) | ("(" +
arithExpr + ")" ) )
relop = oneOf("< == <= >= != <>")

condDef = Forward()
simpleCondExpr = arithExpr + ZeroOrMore( relop + arithExpr ) | condDef
multCondExpr = simpleCondExpr + "*" + arithExpr
condExpr = Forward()
condExpr << ( simpleCondExpr | multCondExpr | "(" + condExpr + ")" )

def reorderArgs(t):
return "cond(" + ",".join(["".join(t.arg3), "".join(t.arg1),
"".join(t.arg2)]) + ")"

condDef << ( Literal("cond") + "(" + Group(condExpr).setResultsName("arg1")
+ "," +
Group(condExpr).setResultsName("arg2")
+ "," +
Group(condExpr).setResultsName("arg3")
+ ")" ).setParseAction( reorderArgs )

tests = [
"cond(a,b,c)",
"cond(1>2,b,c)",
"cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+floa
t(a))",
"cond(a,b,(abs(c) >= d))",
"cond(0,cond(c,cond(e,cond(g,h,(a<f)),(a<d)),(a<b) ),(a<1))",
]

for t in tests:
print t,"->",condExpr.transformString(t)
----------
Prints:
cond(a,b,c) -cond(c,a,b)
cond(1>2,b,c) -cond(c,1>2,b)
cond(-1,1,f)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4 )+(float(b)*2)+float
(a)) ->
cond(f,-1,1)*((float(e)*(2**4))+(float(d)*8)+(float(c)*4)+ (float(b)*2)+float
(a))
cond(a,b,(abs(c) >= d)) -cond((abs(c)>=d),a,b)
cond(0,cond(c,cond(e,cond(g,h,(a<f)),(a<d)),(a<b)) ,(a<1)) ->
cond((a<1),0,cond((a<b),c,cond((a<d),e,cond((a<f), g,h))))
Aug 24 '06 #4
MooMaster Wrote:
I'm trying to develop a little script that does some string
manipulation. I have some few hundred strings that currently look like
this:
cond(a,b,c)
and I want them to look like this:
cond(c,a,b)
I zoned out on your question and created a very simple flipper.
Although it will not solve your problem maybe someone looking for a
simpler version may find it useful as a starting point. I hope it
proves useful. I'll post my simple flipper here:

s = 'cond(1,savv(grave(3,2,1),y,x),maxx(c,b,a),0)'
def argFlipper(s):
''' take a string of arguments and reverse'em e.g.
>>cond(1,savv(grave(3,2,1),y,x),maxx(c,b,a),0)
-cond(0,maxx(a,b,c),savv(x,y,grave(1,2,3)),1)

'''

count = 0
keyholder = {}
while 1:
if s.find('(') 0:
count += 1
value = '%sph' + '%d' % count
tempstring = [x for x in s]
startindex = s.rfind('(')
limitindex = s.find(')', startindex)
argtarget = s[startindex + 1:limitindex].split(',')
argreversed = ','.join(reversed(argtarget))
keyholder[value] = '(' + argreversed + ')'
tempstring[startindex:limitindex + 1] = value
s = ''.join(tempstring)
else:
while count and keyholder:
s = s.replace(value, keyholder[value])
count -= 1
value = '%sph' + '%d' % count
return s

print argFlipper(s)

Aug 24 '06 #5

This thread has been closed and replies have been disabled. Please start a new discussion.