 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Math Forum / Math Software / MATLAB / July 2008inverse regexp
I was wondering if anyone wrote or met a function that was inverting regexp. For instance, let's suppose I want to extract tokens out a a string. I can do that with regexp: >> str = 'Johan and Mary'; >> rex = '(?<husband>\S+) and (?<wife>\S+)'; >> s = regexp(str,rex,'names') s = husband: 'Johan' wife: 'Mary' Now I would like to do the inverse transformation, and have a function that would do this: n = iregexp(rex,s) and would output n = 'Johan and Mary' Has anyone ever met something like this? Or would have an idea on how to do it? jy >>> str = 'Johan and Mary'; >>> rex = '(?<husband>\S+) and (?<wife>\S+)'; >Now I would like to do the inverse transformation, and have >a function that would do this: >n = iregexp(rex,s) >and would output >n = 'Johan and Mary' This is impossible to do in the general case: the regular expressions to be matched upon can include dynamic decision elements that might (for example) base their decisions upon random numbers. Even not including such elements, sub-matches and backwards references make this fairly complicated. For example, the regex might have been (?<husband>\S+([aeiou])\S+) and (?<wife>\S+$2\S+) meaning that pair is only to match if the names share the same vowel. It doesn't take much effort to make this more complicated such that the proper reconstruction of the strings would require a pattern-aware state machine. If you are willing to restrict to very simple patterns, then you can build the necessary routine relatively easily with a few string replacements. Well, except for complications such as if someone decided to make trouble and entered patterns right into what was supposed to be the plain-text string str...  Signature"I was very young in those days, but I was also rather dim." -- Christopher Priest roberson@ibd.nrc-cnrc.gc.ca (Walter Roberson) wrote in message <g6d4al$mij$1@canopus.cc.umanitoba.ca>... > If you are willing to restrict to very simple patterns, then > you can build the necessary routine relatively easily with a few > string replacements. Well, except for complications such as if > someone decided to make trouble and entered patterns right into > what was supposed to be the plain-text string str... Okay, so we can restrict the regexp string to contain only named tokens like in the example above, and nothing else. I have tried ot do it using regexo itself, and to parse the initial regexp string. i was unsuccessful so far, for it only takes the larger possible string out of it, instead of multiple small ones. So you have any tip fr this problem? Cheers jy On Jul 25, 12:33 pm, "Jean-Yves Tinevez" <tinevez.spamprotect...@mpi- cbg.de> wrote: > rober...@ibd.nrc-cnrc.gc.ca (Walter Roberson) wrote in > message <g6d4al$mi...@canopus.cc.umanitoba.ca>... [quoted text clipped - 19 lines] > Cheers > jy I'm not quite sure whether this will help you but look at the 'split' option in the regexp documentation. The example that combines both the 'match' and 'split' options illustrates how to reconstruct the original string. I think this is a relatively new option, I'm using R2007b. HTH, Ashish. I don't know what the ultimate problem that you are trying solve is. But I am guessing that since you are using regular expressions with the token option you are solving a parsing problem. And since you are interested in reversing the parsing you are probably trying to solve a text generation problem. I am wondering if the real problem that you are trying to solve ia a translation problem. If that is the case I would suggest looking at template engines as a way to rewrite the information contained in your tokens. If you need more information on this idea let me know. Donn Thank you all for these very interesting answers! Here is why i am trying to do it: The goal I am pursuing here is just a portable filename parsing system. Trying to analyze the data of a biological screen, we have about 5e4 images to analyze. Each of them as a naming scheme made of a pattern with tokens. For instance: <run_nbr>_YI<chromosome> - <plate_nbr>-<row><col> - f<field> - c<channel>.tiff Using regexp I can get all of the token values automatically. Now I am trying to do the converse problem, that is: trying to rebuild the filename using just the regexp filter and the tokens, so that for instance, I could get all filenames for a request such as "get me all field image for this plate, this row, this col..". Hi Jean-Yves, Interesting problem. To solve the simple case, in which there are only named tokens and literal strings, you can use regexprep to reconstruct the match, but as Ashish pointed out, you will need 'split' to recreate the portions of the initial string that are not matched. To make things simpler, you can restrict your pattern to named tokens that do not contain nested parentheses: >> str = 'Johan and Mary'; >> rex = '(?<husband>\S+) and (?<wife>\S+)'; >> s = regexp(str,rex,'names'); >> regexprep(rex, '\(\?<(\w*)>.*?\)', '${s.($1)}') ans = Johan and Mary See how this fails with a pattern using nested parentheses: >> str = 'Johan and Mary'; >> rex = '(?<husband>[a-z]([aeiou](?=[a-z])[^aeiou])*) and (?<wife>\S+)'; >> s = regexpi(str,rex,'names') s = husband: 'Johan' wife: 'Mary' >> regexprep(rex, '\(\?<(\w*)>.*?\)', '${s.($1)}') ans = Johan[^aeiou])*) and Mary To handle nested parentheses, you need to use a recursive pattern: >> str = 'Johan and Mary'; >> rex = '(?<husband>[a-z]([aeiou](?=[a-z])[^aeiou])*) and (?<wife>\S+)'; >> s = regexpi(str,rex,'names'); >> levelN = '\(([^()]|(??@levelN))*\)'; >> regexprep(rex, '\(\?<(\w*)>([^()]|(??@levelN))*\)', '${s.($1)}') ans = Johan and Mary -=>J Jason Breslau <tendiamonds@mathworks.com> wrote in message > >> str = 'Johan and Mary'; > >> rex = '(?<husband>[a-z]([aeiou](?=[a-z])[^aeiou])*) and (?<wife>\S+)'; > >> s = regexpi(str,rex,'names'); > >> levelN = '\(([^()]|(??@levelN))*\)'; > >> regexprep(rex, '\(\?<(\w*)>([^()]|(??@levelN))*\)', '${s.($1)}') Impressive! This kind of code reminds me now why I never get around to understand fully regexp syntax. Bruno "Bruno Luong" <b.luong@fogale.fr.findthecountry> wrote in message <g6eq9v$cii$1@fred.mathworks.com>... > Impressive! This kind of code reminds me now why I never get > around to understand fully regexp syntax. > > Bruno I'm sorry if I turned you off of it. Regular Expressions are really fun, and can be useful, too, if used properly. The recursive example I used was modified from an example in Jeffrey Friedl's "Mastering Regular Expressions", which is a fantastic book on regular expressions. I couldn't recommend it more: http://regex.info/ -=>J Here is a solution that is a little like swatting flies with a sledge hammer. Download antlrworks from http://www.antlr.org/download and add the jar to your classpath. This will give you access to the stringtemplate template engine. You can use a template to solve your problem with the following code: % create filename template t = org.antlr.stringtemplate.StringTemplate ('$run_nbr$_YI$chromosome$ - $plate_nbr$-$row$$col$ - f$field$ - c$channel$.tiff'); % fake token data run_nbr = '7'; chromosome = 'A2'; plate_nbr = '53'; row = '45'; col = '87'; field = 'myfield'; channel = '45'; % use template to reconstruct filename t.setAttribute('run_nbr', run_nbr); t.setAttribute('chromosome', chromosome); t.setAttribute('plate_nbr', plate_nbr); t.setAttribute('row', row); t.setAttribute('col', col); t.setAttribute('field', field); t.setAttribute('channel', channel); % use filename fileName = char(t); disp(fileName); % reset template to use again t.reset; The first line defining the template is one line. The potential advantage to this approach is that it may be easier to maintain in the future. "Jason Breslau" <tendiamonds@mathworks.com> wrote in message <g6fofo$drn$1@fred.mathworks.com>... > The recursive example I used was modified from an example in > Jeffrey Friedl's "Mastering Regular Expressions", which is a > fantastic book on regular expressions. I couldn't recommend > it more: http://regex.info/ Thank you Jason, I keep the reference in mind. Bruno |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2010 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.