Loading ...
Sorry, an error occurred while loading the content.

Re: [Clip] Re: Line frequency analysis.

Expand Messages
  • John Fitzsimons
    On Mon, 03 Nov 2008 07:01:02 -0500, Don - HtmlFixIt.com wrote: Hi Don, ... Thanks. I did give it a go in 4.95 but unfortunately it didn t finish. It
    Message 1 of 15 , Nov 3, 2008
    • 0 Attachment
      On Mon, 03 Nov 2008 07:01:02 -0500, Don - HtmlFixIt.com wrote:

      Hi Don,

      < snip >

      >Here is how I did something similar without regex -- I bet yours is faster.
      >In my case I am taking one element out of a delimited list vs his
      >example that has just one element, ie, the entire line. Here is mine:

      >:NewTeam
      >;first time set team
      >^!Set %GrabField2%=^$GetField(^$GetRow$;^%TeamField%)$
      >;^!Info ^%GrabField2%
      >^!Set %Team%=^$GetSelection$
      >^!Set %TeamCount%=0
      >;^!Info ^%TeamCount%
      >:Loop
      >;get team for this line
      >^!Set %GrabField2%=^$GetField(^$GetRow$;^%TeamField%)$
      >;be sure that this line is for current team
      >;if not, go to ProcessTeam
      >;otherwise continue here
      >^!If "^%Team%" <> "^$GetSelection$" ProcessTeam
      >^!Set %TeamCount%=^$Calc(^%TeamCount%+1;0)$
      >;^!Info ^%TeamCount%
      >^!Jump +1
      >^!GoTo Loop

      >:ProcessTeam
      >;There is more here that does something with the info
      >^!GoTo NewTeam

      Thanks. I did give it a go in 4.95 but unfortunately it didn't finish.
      It took about a half hour to go through my text file and then kept
      doing something (without producing any output) for another half
      an hour. Before I stopped it.

      Regards, John.
    • Sheri
      ... Can you explain the differences? Might there be spaces at the end of some lines or something? Regards, Sheri
      Message 2 of 15 , Nov 4, 2008
      • 0 Attachment
        --- In ntb-clips@yahoogroups.com, John Fitzsimons <johnf@...> wrote:

        > I did a quick check of some of the totals and although they didn't all
        > match the "Occurrences" I did with S/R they were very close.
        > Certainly fine for what I wanted.

        Can you explain the differences? Might there be spaces at the end of
        some lines or something?

        Regards,
        Sheri
      • Flo
        ... Sheri, Just another idea concerning your solution: In the past, we have used that pattern... ^(.+ r n) 1* quite often for finding duplicate lines. I think
        Message 3 of 15 , Nov 5, 2008
        • 0 Attachment
          --- In ntb-clips@yahoogroups.com, "Sheri" <silvermoonwoman@...> wrote:
          >
          > --- In ntb-clips@yahoogroups.com, John Fitzsimons <johnf@> wrote:
          >
          > >
          > > I want to end up with a list like......
          > >
          > > 000001,0.verizon.windows2000
          > > 000003,0.verizon.windowsxp
          > > 000012,24hoursupport.helpdesk
          > > 000008,alt.computer
          > >
          > > Is there an existing way/clip to do this ? If not then can someone
          > > provide the needed code to produce this result please ?
          > >
          >
          > This will do it exactly as above, but version 5+ is required:
          >
          > ^!SetScreenUpdate Off
          > ^!Jump Doc_End
          > ^!If ^$GetCol$>1 Next Else Skip
          > ^!InsertText ^P
          > ^!Jump Doc_Start
          > :Loop
          > ^!Find "^(.+\r\n)\1*" RS
          > ^!IfError Quit
          > ^!Set %count%=^$StrCount("^%NL%";"^$GetSelection$";Yes;Yes)$
          > ^!Set %fill%=^$Calc(6-^$StrSize(^%count%)$)$
          > ^!Replace "(.+\r\n)\1*" >> "^$StrFill("0";^%fill%)$^%count%,$1" RHS
          > ^!Goto Loop
          > :Quit
          > ^!ClearVariable %count%
          > ^!ClearVariable %fill%
          > ;end of clip


          Sheri,

          Just another idea concerning your solution:

          In the past, we have used that pattern...

          ^(.+\r\n)\1*

          quite often for finding duplicate lines. I think we could also use...

          ^(.+)(\r\n\1)*

          The advantage is that, with this pattern, we don't have to care for
          CRNL at the end of the list.

          In your clip, a final NL is (also) needed for counting the selected
          (duplicate) lines. But, instead of calculating the NL, we could
          write...

          ^$Calc(1+^$GetRowEnd$-^$GetRowStart$)$

          So I think the clip could be slightly shortened like this...


          ^!Jump Doc_Start
          :Loop
          ^!Find "^(.+)(\r\n\1)*" RS
          ^!IfError End
          ^!Set %Count%=^$Calc(1+^$GetRowEnd$-^$GetRowStart$)$
          ^!Set %Fill%=^$Calc(6-^$StrSize(^%Count%)$)$
          ^!Replace "(.+)(\r\n\1)*" >> "^$StrFill("0";^%Fill%)$^%Count%,$1" RS
          ^!Goto Loop

          The only minor disadvantage: Empty lines within the list will provide
          wrong results. So if there are any, we have to remove them with an
          additional command line.

          Do you agree with this solution?

          Regards,
          Flo
           
        • Sheri
          ... That works fine. Here s one that ignores trailing white space and empty lines in the data (but goes back to counting strings for count): ^!SetScreenUpdate
          Message 4 of 15 , Nov 5, 2008
          • 0 Attachment
            --- In ntb-clips@yahoogroups.com, "Flo" <flo.gehrke@...> wrote:
            >
            > Sheri,
            >
            > Just another idea concerning your solution:
            >
            > In the past, we have used that pattern...
            >
            > ^(.+\r\n)\1*
            >
            > quite often for finding duplicate lines. I think we could also use...
            >
            > ^(.+)(\r\n\1)*
            >
            > The advantage is that, with this pattern, we don't have to care
            > for CRNL at the end of the list. In your clip, a final NL is
            > (also) needed for counting the selected (duplicate) lines. But,
            > instead of calculating the NL, we could write...
            >
            > ^$Calc(1+^$GetRowEnd$-^$GetRowStart$)$
            >
            > So I think the clip could be slightly shortened like this...
            >
            >
            > ^!Jump Doc_Start
            > :Loop
            > ^!Find "^(.+)(\r\n\1)*" RS
            > ^!IfError End
            > ^!Set %Count%=^$Calc(1+^$GetRowEnd$-^$GetRowStart$)$
            > ^!Set %Fill%=^$Calc(6-^$StrSize(^%Count%)$)$
            > ^!Replace "(.+)(\r\n\1)*" >> "^$StrFill("0";^%Fill%)$^%Count%,$1" RS
            > ^!Goto Loop
            >
            > The only minor disadvantage: Empty lines within the list will
            > provide wrong results. So if there are any, we have to remove
            > them with an additional command line.
            >
            > Do you agree with this solution?
            >
            > Regards,
            > Flo
            >  
            >

            That works fine. Here's one that ignores trailing white space and
            empty lines in the data (but goes back to counting strings for count):

            ^!SetScreenUpdate off
            ^!Jump Doc_Start
            :Loop
            ^!Find "^(.+)(\s*\r\n\1)*" RS
            ^!IfError Out
            ^!SetArray %farray%=^$GetReSubstrings$
            ;begin long line
            ^!Set
            %Count%=^$StrCount("^%NL%";"^$GetDocReplaceAll("\s*(\r\n)+|(?<=.)\z";"\r\n")$";YES;YES)$
            ;end long line
            ^!Set %Fill%=^$Calc(6-^$StrSize(^%Count%)$)$
            ^!InsertText ^$StrFill("0";^%fill%)$^%Count%,^%farray1%
            ^!Goto Loop
            :Out
            ^!Set %farray%=""
            ^!ClearVariable %farray%
            ^!ClearVariable %Count%
            ^!ClearVariable %Fill%
            ;end of clip
          • Flo
            ... z ; r n )$ ;YES;YES)$ ... Sheri, I think there s a problem with this solution. It ignores trailing blanks and empty lines in the counting of duplicate
            Message 5 of 15 , Nov 6, 2008
            • 0 Attachment
              --- In ntb-clips@yahoogroups.com, "Sheri" <silvermoonwoman@...> wrote:
              >
              > That works fine. Here's one that ignores trailing white space and
              > empty lines in the data (but goes back to counting strings for
              count):
              >
              > ^!SetScreenUpdate off
              > ^!Jump Doc_Start
              > :Loop
              > ^!Find "^(.+)(\s*\r\n\1)*" RS
              > ^!IfError Out
              > ^!SetArray %farray%=^$GetReSubstrings$
              > ;begin long line
              > ^!Set
              > %Count%=^$StrCount("^%NL%";"^$GetDocReplaceAll("\s*(\r\n)+|(?<=.)
              \z";"\r\n")$";YES;YES)$
              > ;end long line
              > ^!Set %Fill%=^$Calc(6-^$StrSize(^%Count%)$)$
              > ^!InsertText ^$StrFill("0";^%fill%)$^%Count%,^%farray1%
              > ^!Goto Loop
              > :Out
              > ^!Set %farray%=""
              > ^!ClearVariable %farray%
              > ^!ClearVariable %Count%
              > ^!ClearVariable %Fill%
              > ;end of clip

              Sheri,

              I think there's a problem with this solution. It ignores trailing
              blanks and empty lines in the counting of duplicate lines but not in
              the ^!Find command. So if we have got...

              0.verizon.windowsxp <-- trailing blank
              0.verizon.windowsxp
              0.verizon.windowsxp

              for example, the clip doesn't find three duplicates but interprets
              this as a singular line plus two duplicates of another line. That is,
              it works fine only on the condition that all three duplicates end
              with a trailing blank.

              So it may be better to remove trailing blanks prior to ^!Find. Isn't
              it?

              Regards,
              Flo

              P.S. By the way: It makes no difference -- but what about...

              ^$GetDocReplaceAll("\s+$|(?<=.)\z";"\r\n")$

              (since NT v.5.0 the \s matches "any white space", including CRNL).
               
            • Sheri
              ... You re right. ... Not necessary if we change the ^!Find to this: ^!Find ^(.+ S)( s* r n 1)* RS ... Haven t tested that and it may work fine. However it
              Message 6 of 15 , Nov 6, 2008
              • 0 Attachment
                Flo wrote:
                > Sheri,
                >
                > I think there's a problem with this solution. It ignores trailing
                > blanks and empty lines in the counting of duplicate lines but not in
                > the ^!Find command. So if we have got...
                >
                > 0.verizon.windowsxp <-- trailing blank
                > 0.verizon.windowsxp
                > 0.verizon.windowsxp
                >
                > for example, the clip doesn't find three duplicates but interprets
                > this as a singular line plus two duplicates of another line. That is,
                > it works fine only on the condition that all three duplicates end
                > with a trailing blank.
                >
                You're right.
                > So it may be better to remove trailing blanks prior to ^!Find. Isn't
                > it?
                >
                Not necessary if we change the ^!Find to this:

                ^!Find "^(.+\S)(\s*\r\n\1)*" RS

                > P.S. By the way: It makes no difference -- but what about...
                >
                > ^$GetDocReplaceAll("\s+$|(?<=.)\z";"\r\n")$
                >
                > (since NT v.5.0 the \s matches "any white space", including CRNL).
                >
                Haven't tested that and it may work fine. However it looks suspicious
                because: we are using the PCRE multiline option by default in NoteTab so
                $ matches at line ends (and line ends by definition are followed by \r\n
                when showing in a NoteTab document) but \s+ matches across line breaks.
                So how can it match $ if it picks up all the \r\n's? In order to match $
                it would need to backtrack. It might create an issue if it backtracks
                between the \r and \n.

                I'll try to test it later.

                Regards,
                Sheri
              • Sheri
                ... Hmn, maybe it should be ^!Find ^(.* S)( s* r n 1)* RS just in case there is only one visible character on the line. Regards, Sheri
                Message 7 of 15 , Nov 6, 2008
                • 0 Attachment
                  Sheri wrote:
                  >
                  > Not necessary if we change the ^!Find to this:
                  >
                  > ^!Find "^(.+\S)(\s*\r\n\1)*" RS
                  >
                  Hmn, maybe it should be

                  ^!Find "^(.*\S)(\s*\r\n\1)*" RS

                  just in case there is only one visible character on the line.

                  Regards,
                  Sheri
                • Sheri
                  ... Indeed that pattern matches in the middle of r n. So after replacement, we end up with r n n. The reason it makes no difference to the outcome is because
                  Message 8 of 15 , Nov 7, 2008
                  • 0 Attachment
                    --- In ntb-clips@yahoogroups.com, Sheri <silvermoonwoman@...> wrote:
                    >
                    > Flo wrote:
                    > > Sheri,
                    > >
                    > > I think there's a problem with this solution. It ignores trailing
                    > > blanks and empty lines in the counting of duplicate lines but not in
                    > > the ^!Find command. So if we have got...
                    > >
                    > > 0.verizon.windowsxp <-- trailing blank
                    > > 0.verizon.windowsxp
                    > > 0.verizon.windowsxp
                    > >
                    > > for example, the clip doesn't find three duplicates but
                    > > interprets this as a singular line plus two duplicates of another
                    > > line. That is, it works fine only on the condition that all three
                    > > duplicates end with a trailing blank. You're right. So it may be
                    > > better to remove trailing blanks prior to ^!Find. Isn't it?

                    > >
                    > Not necessary if we change the ^!Find to this:
                    >
                    > ^!Find "^(.+\S)(\s*\r\n\1)*" RS
                    >
                    > > P.S. By the way: It makes no difference -- but what about...
                    > >
                    > > ^$GetDocReplaceAll("\s+$|(?<=.)\z";"\r\n")$
                    > >
                    > > (since NT v.5.0 the \s matches "any white space", including
                    > > CRNL). Haven't tested that and it may work fine. However it looks
                    > > suspicious because: we are using the PCRE multiline option by
                    > > default in NoteTab so $ matches at line ends (and line ends by
                    > > definition are followed by \r\n when showing in a NoteTab
                    > > document) but \s+ matches across line breaks. So how can it match
                    > > $ if it picks up all the \r\n's? In order to match $ it would
                    > > need to backtrack. It might create an issue if it backtracks
                    > > between the \r and \n.
                    >
                    > I'll try to test it later.

                    Indeed that pattern matches in the middle of \r\n. So after
                    replacement, we end up with \r\n\n. The reason it makes no difference
                    to the outcome is because we are counting "\r\n" after applying this
                    in ^$GetDocReplaceAll$. You would be able to see a problem if that
                    were getting inserted in the document (however, something else happens
                    when you insert it, \r\n\n becomes \r\n\r\n because the input control
                    needs line breaks to be \r\n -- you can test the string size prior to
                    insertion vs testing the size of a selection after insertion).

                    I think it is preferable to greedily replace white space that precedes
                    "\r\n" with "". The white space does include other CRLFs.

                    Regards,
                    Sheri
                  • Flo
                    ... Yes, I can see now what happens here. I think your explanation is in better accordance with the PCRE Documentation than the NoteTab Help on RegEx. The
                    Message 9 of 15 , Nov 8, 2008
                    • 0 Attachment
                      --- In ntb-clips@yahoogroups.com, "Sheri" <silvermoonwoman@...> wrote:
                      >
                      > Indeed that pattern (Flo:^$GetDocReplaceAll("\s+|(?<=.)\z";"\r\n")$
                      > matches in the middle of \r\n. So after replacement, we end up with
                      > \r\n\n...The reason it makes no difference...

                      Yes, I can see now what happens here. I think your explanation is in
                      better accordance with the PCRE Documentation than the NoteTab Help
                      on RegEx. The latter says: "$ assert end of string (or line, in
                      multiline mode)".

                      The PCRESYNTAX Documentation from PCRE 7.7 is more detailed: "$ end
                      of subject, also before newline at end of subject, also before
                      internal newline in multiline mode."

                      Thanks again, Sheri! I can clearly see the difference now and
                      why "^$GetDocReplaceAll("\s+|(?<=.)\z";"\r\n")$" doesn't affect the
                      result.

                      Flo
                       
                    Your message has been successfully submitted and would be delivered to recipients shortly.