sort file removing duplicates possible?

Scott Sumner

Your regular expression will not work as shown; I can tell some of the important characters in it are missing (due to being consumed by the Markdown syntax this website uses). If you need to write a * here, either escape it, like this: \* in which it will then appear like this: *…or put it inside backticks (also called grave accents), like this `*` in which case it will then appear like this: * (I find this second way of doing it easier/cleaner).

So let’s assume you meant to post ^(.*?)$\s+?^(?=.*^\1$)

That is clearly wrong as it introduces the notion of whitespace (with the \s+). That may work for exactly what YOU are doing, but in general; if we are sorting lines and then deleting duplicates then \s doesn’t belong.

I retested the original test sort and del dupe lines macro definition I posted earlier, and it (still) works. So let’s try an example, here’s some random unsorted data with some duplicates:

nonprehensile
aeriferous
calabrian
estimating
urethrostomy
redigging
estimating
crystalling
cynwulf
fetor
kali
armada
auspicate
tusser
fetor
zoophyte
ribless
fetor
deducibly
calabrian
estimating
expressivity

After running the macro, we obtain the desired result (sorted, duplicates removed), nothing more, nothing less:

aeriferous
armada
auspicate
calabrian
crystalling
cynwulf
deducibly
estimating
expressivity
fetor
kali
nonprehensile
redigging
ribless
tusser
urethrostomy
zoophyte

Earlier I presented the regular expression (in the macro) without explanation. Here’s what it is actually doing to find matches:

(?-s)   ^   (.*)   (?:   \R   )   (?s)   (?=   .*   ^   \1   \R   )
^^^^^-----------------------------------------------------------------------------> after this . appearing will match any single character except a line-ending character
        ^-------------------------------------------------------------------------> match only at start of line
            ^  ^------------------------------------------------------------------> remember what matches the expression between `(` and `)`; call it "group #1"
             ^^-------------------------------------------------------------------> match zero or more characters (in this case it will stop matching at the next line-ending character)
                   ^^^        ^---------------------------------------------------> the syntax `(?:` followed by something followed by `)` means to match what is inside but don't remember it
                         ^^-------------------------------------------------------> match a line-ending
                                  ^^^^--------------------------------------------> after this a . appearing will match any single character including a line-ending character
                                         ^^^                      ^---------------> the syntax `(?=` followed by something followed by `)` means whatever is inside needs to match for the overall expression to match, but it isn't consumed and isn't part of the match
                                               ^^---------------------------------> match zero or more characters (ANY characters, including line-ending characters...this will match across lines)
                                                    ^-----------------------------> match at start of line
                                                        ^^------------------------> match what we remembered earlier as "group #1"
                                                             ^^-------------------> match a line-ending character

So that’s actually still pretty technical/confusing. Maybe this is better:

look at the complete contents of the current line
glance further down (any arbitrary amount of distance) in the file to see if there is an exact duplicate of the current line contents; if so, the current line is a MATCH
if there was a MATCH, do the REPLACEMENT (see below)
move beyond the current line (go down one line)
repeat these steps until the file is exhausted

When there is a MATCH, perform the specified REPLACEMENT on the matched text. Since the matched text is the entire current line (with line-ending), and the replacement is NOTHING, this effectively deletes the current line. This is what is desired because our goal is to removed duplicate lines and we’ve already established that the matched line occurs again further down in the file.

Dedicated regexers might catch that my regular expression is more-complicated than needed. The (?:\R) part could be changed to simply \R. (?: ) is normally used when some grouping without remembering is necessary. Here \R needs no grouping so the (?: ) is just extra fluff…does no harm except making the whole expression harder to read.

Super-dedicated regexers would also say that the regular expression is more-complicated than needed because at the point it is run (inside the macro above) the file is already sorted! Very true. The regular expression above works for sorted or unsorted duplicate line detection/removal. Here’s the simpler one that will only work on the sorted lines situation: (?-s)^(.*)\R(?=^\1\R)

One more thing worth mentioning is that if you want to delete duplicate lines but ignore/leave blank/(empty!) lines, simply change (.*) in the regular expression to (.+). Instead of “match ZERO or more characters”, this change will “match ONE or more characters”, effectively failing to match truly empty lines so they won’t be removed.

patrickdrd

I had problem with a file today, I don’t know why,
is there any way I can send you the file to take a look?

Scott Sumner

@patrickdrd said:

send you the file to take a look

Post its contents on http://textuploader.com/ and reply here with the link to it. Also indicate exactly what problem you are seeing if it isn’t obvious. No guarantees about how deeply I can get involved, but I’ll take a quick look… :-)

patrickdrd

ok, it’s here:
http://textuploader.com/d4dkr

removing duplicates results contain a duplicate result

patrickdrd

remove duplicates without sorting produces the result:
DBServerIP=10.1.249.215
DBServerIP=10.12.77.185
DBServerIP=10.1.249.215

(1st and 3rd exact duplicates)

and sorting first and then removing dups results in:
DBServerIP=10.12.22.129
DBServerIP=10.12.77.185
DBServerIP=10.12.77.185

last 2 duplicated

(using your regular expression in both cases of course)

Scott Sumner

@patrickdrd

Okay, so a quick look told me what is happening. It’s interesting. :-)

If you run only the sort on this data, you’ll get this at the end of the file:

Imgur

You’d think that the removal of duplicate lines after this would result in only a single occurrence of DBServerIP=10.12.77.185. However, if we look closer at what is really left after the duplicate removal (turning on visibility of line-endings!), we see:

Imgur

We see that the two lines with that IP address truly are different–one has a line-ending and one doesn’t–and because these two lines are not the same, the regular-expression replacement is working correctly by leaving both of these lines after it does its work.

All of the original lines that ended in .185 had line-endings (in other words, were exactly the same), so I’d say this is an artifact resulting from the sort operation (in my mind this is a sort BUG!).

But we can work around it. There could be a regular expression solution, but maybe the regex that removes duplicates is complicated enough. What I’d suggest here is to modify the original macro, after the sort but before the find+replace, to:

move caret position to the end of file (the sort operation leaves it at the beginning of file)
insert a (Windows style) line-ending
move caret position back to beginning of file (in preparation for the find+replace)

Thus:

<Macro name="test sort and del dupe lines 2" Ctrl="no" Alt="no" Shift="no" Key="0">
    <Action type="2" message="0" wParam="42059" lParam="0" sParam="" />
    <Action type="0" message="2318" wParam="0" lParam="0" sParam="" />
    <Action type="1" message="2170" wParam="0" lParam="0" sParam="&#x000D;" />
    <Action type="1" message="2170" wParam="0" lParam="0" sParam="&#x000A;" />
    <Action type="0" message="2316" wParam="0" lParam="0" sParam="" />
    <Action type="3" message="1700" wParam="0" lParam="0" sParam="" />
    <Action type="3" message="1601" wParam="0" lParam="0" sParam="(?-s)^(.*)(?:\R)(?s)(?=.*^\1\R)" />
    <Action type="3" message="1625" wParam="0" lParam="2" sParam="" />
    <Action type="3" message="1602" wParam="0" lParam="0" sParam="" />
    <Action type="3" message="1702" wParam="0" lParam="512" sParam="" />
    <Action type="3" message="1701" wParam="0" lParam="1609" sParam="" />
</Macro>

Compare that with the earlier version much earlier in this thread.

Running this new macro on your data results in:

Imgur

which is the desired result.

Scott Sumner

@patrickdrd

For the DBServerIP=10.1.249.215 case you mentioned, if you don’t do the sort first, then a .215 line is the last line of the data you turn over to the regular expression replace operation. Thus it lacks the trailing line-ending that the earlier occurrence of a similar line has. Same issue as the .185 case…

patrickdrd

thanks a ton!

I’d a like a more “generic” approach, so I called trim 42056 to clear empty lines first,
because sorting puts an empty line at the top if is finds one,
then going on as you suggested

Scott Sumner

@patrickdrd

I’m glad you have a solution. Yeah, the empty line thing with sorting is rather bad, but the end-user can’t control this behavior of the sorting. I think I’ve changed my mind and it is probably best to alter the regular expression a bit in order to handle the situation where there is a duplicate-but-without-line-ending at the end of the file. So I’d suggest changing it to:

(?-s)^(.*)\R(?s)(?=.*^\1(?:\R|\z))

I’ve done two things to this regex:

I removed the (?: and ) around the first \R (a simplification discussed earlier so no need to say any more here)
The final \R was changed to (?:\R|\z) (see discussion below)

The \R|\z part is what allows an almost-duplicate at the end-of-file-without-line-ending to be detected. The new part to this is the \z which roughly means “match only at the very end of the data”.

The (?: and ) was added so that the | only affects the \R that precedes it and the \z that follows it

guy038

Hello @patrickdrd, @scott-sumner and All,

Generally speaking, when you want to remove duplicate lines, from a PREVIOUSLY SORTED list, just use this simple regex S/R, below :

SEARCH (?-s)(.*\R)\1+

REPLACE \1

This regex is quite fast, because, in case of numerous duplicates, the part \1+ grabs all the duplicates ( with their EOL characters ), at once and just rewrites the first item of each block :-))

IMPORTANT : the last item of your sorted list must be followed by EOL character(s) !

Cheers,

guy038

patrickdrd

after trying both the macro and the textfx solution for a long time,
I’ve seen that still ultraedit’s sorting works much better than both of them,
I first rejected the macro for textfx’s favor but I found out lately that neither the latter does a good job, at least I prefer the sorting done by ultraedit ,sorry,
I don’t have an example at the moment, I’ll post again when I do

Claudia Frank

@patrickdrd

sorry, did not read the whole thread but what about a python one liner?

editor.setText('\r\n'.join(list(set(editor.getText().splitlines()))))

Cheers
Claudia

Claudia Frank

@patrickdrd

forgot sorting

editor.setText('\r\n'.join(sorted(list(set(editor.getText().splitlines())))))

Cheers
Claudia

patrickdrd

thanks, I’m keeping that too and I’ll let you know

patrickdrd

ok, I tested, this doesn’t work either somehow, at least doesn’t work like ue’s one, i.e.

I tested with easylist from: https://easylist.to/easylist/easylist.txt (69883 lines),
ue’s function cuts it to: 69238

npp results:
python scipt: 69818
macro (scott) is a bit slow and results in 19610 lines (!) and
textfx results in 69818 as well

guy038 regular expression results in 28109 lines,
according to my experience, I bet that ue’s result is the correct one,
at least to my taste

Claudia Frank

@patrickdrd

could you, by any chance, upload the ue cutted list?
To see the differences.

Cheers
Claudia

patrickdrd

yes, of course, please tell me where, pastebin doesn’t work, it’s blocked here (at work),
any other suggestions?

Claudia Frank

actually pastebin is my first choice as well and haven’t used others for quite some time now.

Heard about

https://www.zippyshare.com/
https://www.sendspace.com/

should be good and anonymous but haven’t tried it so far.

Cheers
Claudia

Scott Sumner

Yea, I used to be a fan of regular expression replacement when doing this, but with “larger” datasets there always seems like there is so much tweaking and experimentation needed to get it right (for a particular dataset) that it is hardly worth it, unless you like playing with regular expressions all day instead of solving a particular problem and moving on quickly.

A Pythonscript solution such as @Claudia-Frank 's seems fine…

patrickdrd

ok, but why results are inconsistent (with large datasets)?

@Claudia-Frank, unfortunately I’m not able to access zippyshare either,
so I’ll upload to pastebin from home if we don’t find another solution