Reg Ex to find lines with linefeeds in csv

Mark Yorkovich · Sep 6, 2019, 3:55 PM

Thanks for the reply, Alan. A “multistep approach” isn’t ideal in this situation. The file is over 100k lines - something like 70k records and I’d like to have a one-step process where I enter the regex in the Find What box, set my replacement string and click Replace All. I’ve been manually going through the file, fixing the offending records, but that way will take an incredible amount of time.

I see what you’re intending to do, which made me realize that there can be characters in the last column - after the last pipe and before the LF. I failed to mention that in my post.

Alan Kilborn · Sep 6, 2019, 5:53 PM

@Mark-Yorkovich

A “multistep approach” isn’t ideal in this situation

It seems like my 3 steps could all be of the Replace All variety, thus making it only 3 actions for a file of arbitrary length. Not 3 steps you would have to repeat n times.

there can be characters in the last column - after the last pipe and before the LF. I failed to mention that in my post.

Actually, I think you did mention it, indirectly, by showing these lines, which have that attribute:

lorem|1|2|1|1|1|0|||9887544[LF]
dolor|6|6|1|1|0|0|||123456[LF]

Mark Yorkovich · Sep 6, 2019, 5:54 PM

Okay - I’ll pull a bunch of lines out of the file into a test file and give it a whirl.

Alan Kilborn · Sep 6, 2019, 6:15 PM

@Mark-Yorkovich

I still think the best approach is to temporarily change the [LF] you want to keep into something else, then go back and change any [LF] remaining (the bogus ones) into what you want them to be. And finally revert the substituted line-ends back into real line-ends.

Note that I don’t think there is a good way to detect a bogus [LF] if it occurs in the final field (after the 9th pipe), so I hope that doesn’t occur for you.

This is rather a tough problem because the pattern can get destroyed quickly.

Mark Yorkovich · Sep 6, 2019, 6:15 PM

So, here’s a good example of what I’m working with. The multi-line records are the lines with asterisks by the line number. This forum text makes it kinda hard to read like this, but hopefully you’ll get the idea.

1 lorem62|8|5|1|1|0|0|||EQUI033[LF]
2 lorem62|8|6|1|1|0|0|||EQUI036[LF]
*3 lorem62|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit[LF]
*4 quis nostrud exercitation ullamco[LF]
*5 laboris nisi ut aliquip ex ea commodo consequat.[LF]
*6 consectetur adipiscing elit|[LF]
7 lorem-44|1|1|1|1|0|0|||3003024[LF]
8 lorem-44|1|2|1|1|0|0|||3001476[LF]
9 lorem-44|1|3|1|1|0|0|||2001415[LF]
10 lorem47|8|4|1|1|0|0|||EQUI032[LF]
11 lorem47|8|5|1|1|0|0|||EQUI033[LF]
12 lorem47|8|6|1|1|0|0|||EQUI036[LF]
13 lorem47|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit|[LF]
14 lorem62|1|1|1|1|0|0|||3003024[LF]
15 lorem62|1|2|1|1|0|0|||3001476[LF]
16 lorem-44|8|6|1|1|0|0|||EQUI036[LF]
*17 lorem-44|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit[LF]
*18 quis nostrud exercitation ullamco[LF]
*19 laboris nisi ut aliquip ex ea commodo consequat.[LF]
*20 consectetur adipiscing elit[LF]
*21 |[LF]
22 lorem20|1|1|1|1|0|0|||3003024[LF]
23 lorem20|1|2|1|1|0|0|||3001476[LF]
24 lorem20|1|3|1|2|0|0|||3001894[LF]
25 lorem20|1|4|1|1|0|0|||2001443[LF]
26 lorem20|1|5|1|1|0|0|||2001439[LF]
27 lorem20|1|6|1|1|0|0|||3002274[LF]
28 lorem10|5|1|1|1|1|0|||MED0015[LF]
*29 lorem20|13|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit[LF]
*30 quis nostrud exercitation ullamco[LF]
*31 laboris nisi ut aliquip ex ea commodo consequat.[LF]
*32 consectetur adipiscing elit|[LF]
33 lorem20|3|1|1|2|0|0|||3002407[LF]
34 lorem16|11|0|1|0|0|0|||[LF]

So, according to the steps, I should first replace pipe nand line feed |[LF] with [CR] (Find what: \n and Replace with: \r)
I did that and lines 6, 13, 21, 32 and 34 now end with [CR]. Right away, I see this won’t work, as we’re effectively removing the last column from those lines. Remember, the rows should be 10 columns, with 9 pipes separating the columns.

Mark Yorkovich · Sep 6, 2019, 6:38 PM

Also, step 2 replaces the rest of the [LF] with a dummy string, resulting in 5 long lines of text, with records/lines merged.

Alan Kilborn · Sep 6, 2019, 6:53 PM

The earlier steps were made with some bogus assumptions due to not enough data.

And/or just some plain old mistakes. :)

So taking the new, lengthier data, which in reality is (I think) this, in a Linux formatted N++ tab:

lorem62|8|5|1|1|0|0|||EQUI033
lorem62|8|6|1|1|0|0|||EQUI036
lorem62|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit
quis nostrud exercitation ullamco
laboris nisi ut aliquip ex ea commodo consequat.
consectetur adipiscing elit|
lorem-44|1|1|1|1|0|0|||3003024
lorem-44|1|2|1|1|0|0|||3001476
lorem-44|1|3|1|1|0|0|||2001415
lorem47|8|4|1|1|0|0|||EQUI032
lorem47|8|5|1|1|0|0|||EQUI033
lorem47|8|6|1|1|0|0|||EQUI036
lorem47|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit|
lorem62|1|1|1|1|0|0|||3003024
lorem62|1|2|1|1|0|0|||3001476
lorem-44|8|6|1|1|0|0|||EQUI036
lorem-44|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit
quis nostrud exercitation ullamco
laboris nisi ut aliquip ex ea commodo consequat.
consectetur adipiscing elit
|
lorem20|1|1|1|1|0|0|||3003024
lorem20|1|2|1|1|0|0|||3001476
lorem20|1|3|1|2|0|0|||3001894
lorem20|1|4|1|1|0|0|||2001443
lorem20|1|5|1|1|0|0|||2001439
lorem20|1|6|1|1|0|0|||3002274
lorem10|5|1|1|1|1|0|||MED0015
lorem20|13|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit
quis nostrud exercitation ullamco
laboris nisi ut aliquip ex ea commodo consequat.
consectetur adipiscing elit|
lorem20|3|1|1|2|0|0|||3002407
lorem16|11|0|1|0|0|0|||

The following 3 replace-all transforms seem to do the job:

F: (?-s)^([^|]*?\|){9}.*\K\n
R: \r

F: \n
R: LINEFEED <— your choice of text substitution here

F: \r
R: \n

After which one obtains:

lorem62|8|5|1|1|0|0|||EQUI033
lorem62|8|6|1|1|0|0|||EQUI036
lorem62|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elitLINEFEEDquis nostrud exercitation ullamcoLINEFEEDlaboris nisi ut aliquip ex ea commodo consequat.LINEFEEDconsectetur adipiscing elit|
lorem-44|1|1|1|1|0|0|||3003024
lorem-44|1|2|1|1|0|0|||3001476
lorem-44|1|3|1|1|0|0|||2001415
lorem47|8|4|1|1|0|0|||EQUI032
lorem47|8|5|1|1|0|0|||EQUI033
lorem47|8|6|1|1|0|0|||EQUI036
lorem47|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elit|
lorem62|1|1|1|1|0|0|||3003024
lorem62|1|2|1|1|0|0|||3001476
lorem-44|8|6|1|1|0|0|||EQUI036
lorem-44|9|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elitLINEFEEDquis nostrud exercitation ullamcoLINEFEEDlaboris nisi ut aliquip ex ea commodo consequat.LINEFEEDconsectetur adipiscing elitLINEFEED|
lorem20|1|1|1|1|0|0|||3003024
lorem20|1|2|1|1|0|0|||3001476
lorem20|1|3|1|2|0|0|||3001894
lorem20|1|4|1|1|0|0|||2001443
lorem20|1|5|1|1|0|0|||2001439
lorem20|1|6|1|1|0|0|||3002274
lorem10|5|1|1|1|1|0|||MED0015
lorem20|13|1|1|0|0|0||Lorem ipsum dolor sit amet, consectetur adipiscing elitLINEFEEDquis nostrud exercitation ullamcoLINEFEEDlaboris nisi ut aliquip ex ea commodo consequat.LINEFEEDconsectetur adipiscing elit|
lorem20|3|1|1|2|0|0|||3002407
lorem16|11|0|1|0|0|0|||

Mark Yorkovich · Sep 6, 2019, 7:06 PM

IT WORKED! Actually I stopped at step 2, as I just need to get it to a point where each record is on a single line - the EOL character(s) don’t matter - it just needs to be able to be opened in Excel.

Dude, do you realize how much time you just saved me?! :)

Thanks!

Alan Kilborn · Sep 6, 2019, 7:30 PM

@Mark-Yorkovich said:

Thanks!

No problem. It was an interesting thing to think about. I like some minor challenges on a daily basis to keep the aging little grey cells (hopefully) fine-tuned…

Key to your problem was grabbing enough text such that the real line-ending [LF] could be isolated. Once that’s done, it becomes rather trivial…

guy038 · Oct 19, 2019, 2:05 AM

Hello, @mark-yorkovich, @alan-kilborn and All,

I was away from the N++ community since more than a month ( see my post below )

https://community.notepad-plus-plus.org/post/47895

So, Mark, my reply is quite out of date. However, just for info, here is a one-go regex, which does what you want :

SEARCH (?-s)([^|\n]*\|){9}.*\n|\n(?!([^|\n]*\|){9}.*\n)

REPLACE (?1$0:LINEFEED)

It’s the equivalent regex S/R to the Alan’s one. As you just want to replace with a single dummy char, use, for instance, the following Replace zone :

REPLACE (?1$0:#)

Notes :

First, the part (?-s) forces the regex engine to consider any dot ( . ) as a single standard character ( Not an EOF one ! )
Then, the search regex processes one of the two alternatives :
- (?-s)([^|\n]*\|){9}.*\n which looks for entire correct lines, containing 9 pipe characters. In replacement, as group 1 exists, the overall match ( $0 ) is just re-written
- \n(?!([^|\n]*\|){9}.*\n) which simply searches a line-feed char, only if NOT followed with a correct record. In replacement, as group 1 does not exist, the part after the colon ( LINEFEED or # ) replaces the \n character

Best Regards,

guy038