Find line above given text in document
-
Hi Eko, thank you for quick response, I have managed to sort it now using Python.
Tooltip in screenshot.
No invisible characters by the way, and it works fine with a smaller amount of lines.
-
@Benji2025 said in Find line above given text in document:
The suggested method works fine on a small number of lines but with millions of lines
I tested this with 120_000_000 lines and it worked for me, but to be honest, it only had 1.5 GB, so … your file must be much larger accordingly. Unfortunately I don’t know the internals from what size this becomes a problem.
I have managed to sort it now using Python
hehe … from my point of view that is ALWAYS the solution :-D
-
Hello, @benji2025, @ekopalypse, @alan-kilborn and All,
Oh, My God, I’ve been beaten by @ekopalypse :-((
@benji2025, I’d really like to know the average size of your files and their number of lines !
Indeed, I did a test with a file of size
143,151,374bytes, containing3,151,513lines of47characters each. And, both, my or your regex worked fine and mark121,212lines !!So, your regex that I used is
(?-s)^.*$(?=\RTEST)And I used a similar syntax
(?-s)^.*\R(?=TEST$)
On my old
Win XPmachine, with N++v7.9.2release, :-
Your regex did the marking operation in about
24,3seconds -
My regex did the marking operation in about
23,2seconds
So, I suppose that you should try my regex version !
Best Regards,
guy038
-
-
Thanks guys
Just the 9 million lines at 1GB, and that is just the short run of a job I am running.
Same error with guy038’s syntax.
-
@Ekopalypse said in Find line above given text in document:
I have managed to sort it now using Python
hehe … from my point of view that is ALWAYS the solution :-D
Yeah, to expand on that for the benefit of others who don’t know: Python’s
relibrary is usually at least 10x faster than Notepad++'s built-in search capability, such that it is vastly better when searching extremely large files. The Columns++ and MultiReplace plugins, while very powerful in their own right, will AFAIK never be much faster than the Notepad++ find/replace form because they also do their search-replace operations through Scintilla.@PeterJones
At this point I’ve repeated this PSA enough times that it should probably be added to one of the FAQ’s, maybe this one? -
@Mark-Olson said in Find line above given text in document:
The Columns++ and MultiReplace plugins, while very powerful in their own right, will AFAIK never be much faster than the Notepad++ find/replace form because they also do their search-replace operations through Scintilla.
A minor technical quibble: the regex search in Columns++ search does not use Scintilla search. While it does search within Scintilla’s buffer (avoiding a copy, using a documented interface that exposes the content as addressable bytes), it uses Boost::regex directly, with its own custom iterators, rather than the Scintilla search interface.
Still, this probably only makes it approximately equal in speed relative to Notepad++, since it does use the same regular expression engine in essentially the same way.
-
I’ll have to see if I can test this with some large files, but does it not strike you as very strange that this expression:
(?-s)^.*$(?=\RAccess is denied)should yield a complexity error, regardless of the data? Complexity errors are supposed to happen when the same text keeps getting rescanned; that is, not just when the operation takes a long time, but when the number of bytes examined is growing “too much” faster than the start point is being moved forward (or else when internal stacks overflow preset bounds). This expression shouldn’t cause that. It doesn’t backtrack.
@Benji2025 — I know you’ve solved your problem, but if you are still reading and interested: Does the same thing happen on the same data with:
(?-s)^.*+(?=\RAccess is denied)That shouldn’t matter, but maybe the regex engine isn’t as smart as I think it is.
-
@Mark-Olson said in Find line above given text in document:
At this point I’ve repeated this PSA enough times that it should probably be added to one of the FAQ’s, maybe this one?
The suggestion to “use Python” is not necessarily the same as the suggestion to “use PythonScript”; I am not sure whether @Benji2025 was using the standalone python interpreter, or using the Notepad++ automation plugin to search the open file. I am also not sure whether your comment about performance is saying “
python.exe’srelibrary is faster” or “using PythonScript and therelibrary is faster” or “using PythonScript and it’sre-likeeditor.research()is faster”. Because those are three different things.The FAQ you pointed to is only about using PythonScript plugin (or other plugins) to do mathy-replacements; the generalized statement you seem to be making doesn’t seem to be restricted to mathy-replacements, so I’m not sure that’s the best place, even once the context of the claim is clarified.
So that we don’t clutter this specific question with workshopping, if you wanted to make an RFC post in the Blogs category to workshop a new FAQ entry, once everyone was happy with everything workshopped, I could create a new entry in the FAQ category and duplicate the final version of the post (with you as the author).
-
@Coises said in Find line above given text in document:
I’ll have to see if I can test this with some large files, but does it not strike you as very strange that this expression:
I tried thinking a bit laterally about this issue, more specifically the error of “complexity” stopping the regular expression from completing.
Since we all seemed unsure of why it would generate such a message from what “seemed” to be a simple find expression I thought I would do a small amount of testing to see if the
.*or the lookahead was likely to blame.I created a file with the lines
bla bla bla blaandAccess is deniedon a ratio of about 10 bla to 1 of Access. I got to 240M lines approx (so by my calculation around 3.6Gb) at which point the lags in updating NPP were significant. At this point I gave the original regex a try. Work called so after I completed that job I came back to a very sorry dual monitor Windows 11 system. One monitor had called it quits (Windows wouldn’t use it) and the smaller monitor had all windows squished on it. On the upside, the regex worked, whereas I actually wanted it to fail so I could hatch my next cunning step, a revised regex to continue testing.So back to my idea of seeing which part might be the cause of the issue. Since the original request was to be able to mark the lines I considered not trying to "capture’ the entire line, instead just use something like
.\R(?=Access is denied$). Mark function will still mark the line even if only 1 character on that line is sought.Another similar idea would have been to (again) just select the last character on a line and also select the following line with “access is denied.” If the reason for the “mark” was to extract them to another tab/file, then it would be much simpler to remove the “access is denied” line at that time.
There was also a 3rd idea. That is to remove the
\Rimmediately before the “access is denied” line. Then use the mark function to mark the actual text “Access is denied”. Based on this another slightly modified idea is to copy the “Access is denied” to the end of the line above. Then Mark lines which have this text but not from the start of the line.As shown, there are often many answers/solutions to the problem, especially if one is willing to divide and conquer. It is nice to give a one line solution but often complexities (inability to solve or easily adjust or user unable to comprehend the solution) will make a multiple step solution more palatable.
Terry
-
I wrote:
I’ll have to see if I can test this with some large files
I have a moderately large file (19,473 lines, 119,172,867 bytes) I sometimes use for testing. It does not contain any lines which begin
Access is denied.This expression:
(?-s)^.*$(?=\RAccess is denied)
results in the complexity error message; this one:
(?-s)^.*+(?=\RAccess is denied)
correctly returns zero matches.So the regular expression engine does not optimize the first expression to the equivalent second expression.
-
@Coises said in Find line above given text in document:
So the regular expression engine does not optimize the first expression to the equivalent second expression.
So although the
.*$doesn’t appear to allow for backtracking, maybe that’s what the engine thinks is possible, hence the error. In the second the.*+states "there CANNOT be any backtracking!So if the OP changed the
$to a+that alone might be sufficient to allow for the regex to work.Terry
PS actually the more I think about it, the
.*$will allow for backtracking. My (Our) thinking likely has to change. Although$is a meta-character, it still doesn’t have the “power” to command the engine to not backtrack, whereas the+does. It might even be such that any character at the$position isn’t regarded as an anchor to prevent backtracking when/if deemed (possibly) needed by the engine. I’d be interested in going back over some of these errors reported and seeing if it is possible to add the possessive modifier and re-test. -
@Terry-R said in Find line above given text in document:
Since we all seemed unsure of why it would generate such a message from what “seemed” to be a simple find expression I thought I would do a small amount of testing to see if the .* or the lookahead was likely to blame.
A truly strange result happened when I tried removing the lookahead and using plain old Count in the Find dialog. With my 19,473 line, 119,172,867 byte file, entering either of these expressions:
(?-s)^.*$
(?-s)^.*+
into the Find window and pressing Count causes Notepad++ to hang (“Not Responding”). I’ve waited over six minutes before force closing.So, I tried to Count one of those expressions in the search in my Columns++ plugin, because (depending on the cause) that can show a progress meter for slow operations, and I wanted to get an idea what was happening.
It completed, with the correct answer (19,473, the number of lines) in around one second. The result is the same with either expression. (The original expressions behave as in Notepad++: the version with the dollar sign gets a complexity error, the version with a plus sign works.)
Now given that both use Boost::regex, I have no idea why Notepad++ hangs.
-
@Coises said in Find line above given text in document:
A truly strange result happened when I tried removing the lookahead and using plain old Count in the Find dialog. With my 19,473 line, 119,172,867 byte file, entering either of these expressions:
(?-s)^.*$
(?-s)^.*+
into the Find window and pressing Count causes Notepad++ to hang (“Not Responding”). I’ve waited over six minutes before force closing.Now given that both use Boost::regex, I have no idea why Notepad++ hangs.
Ugh. Doesn’t hang in 8.6.6 portable. Which gives me a bad feeling it could be related to PR #16208.
Edit to add: Reported in 8.7.9 announcement as a regression. Still studying the cause.
-
@guy038 said in Find line above given text in document:
Oh, My God, I’ve been beaten by @ekopalypse :-((
What’s that suppose to mean?? Lol - just kidding :-D
-
Hello, @ekopalypse,
May be, it’s a language barrier ! It was, in no way, offensive to you !
I just wanted to say that your example seemed closer to @benji2025 case and that is was impossible for me to compete with you (
120,000,000 lines!) ;-)))BR
guy038
-
Hi, @benji2025, @ekopalypse, @alan-kilborn, @coises and All,
I did additional tests :
First, if the
Word wrapoption is enabled, using the same regex as before, with theMatch Caseoption checked, the BookMarking operation took about the same time :23.2secondsSecondly, after the BookMarking operation if you re-run the same regex, the whole operation is done in
16seconds. This seems logical because n++ does not have to re-bookmark the already bookmarked lines !Thirdly, if I do not check the
Bookmark lineoption, in the Mark dialog, the Marking operation is a bit quicker :19,4secondsFourthly, if I select all the contents of the test file ( => the
In selectionbox is automatically checked ) the operation is a bit slower :23,8seconds
Now, for all the tests below, I used these rules :
-
The
Word wrapoption is unchecked -
In the Mark dialog, the
Bookmark lineoption is checked and all the other box options are unchecked. -
Generally the
Match caseoption is unchecked but may be checked in few occasions. -
Before each search, I hit the
Clear all marksbutton and place the caret at the very beginning of the test file. -
I did my tests twice : on my old
XPmachine, with N++v7.9.2and on my newW10laptop, withN++ v8.7.6( @coises, I avoided, on purpose, using thev8.7.8andv8.7.9releases ! ) -
Each time, I opened N++, from a command prompt window, with the command
Notepad++ -nosession Benji.txt Test_Benji.txt, so with only these two files. -
For the
W10test, I simply used an USB key containing the portable N++v8.7.6release and the test file.
-------------------------- NEC XP - N++ v7.9.2 ------------------------------------------------------------------------------------------------------- (?-s)^.*\R(?=TEST$) 24 s Option 'Match Case' unchecked (?-s)^.*\R(?=TEST$) 23.1 s Option 'Match Case' checked ( The test in my **previous** post ) (?-s)^.+\R(?=TEST$) 23.9 s Option 'Match Case' unchecked (?-s)^.*+\R(?=TEST$) 19.9 s ( Atomic ) Option 'Match Case' unchecked (?-s)^.++\R(?=TEST$) 19.8 s ( Atomic ) Option 'Match Case' unchecked (?-s)^.++\r\n(?=TEST$) 18.6 s ( Atomic ) Option 'Match Case' unchecked (?-s)^.++\r\n(?=TEST$) 17.7 s ( Atomic ) Option 'Match Case' checked (?-is)^.++\r\n(?=TEST$) 69 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked --- Without the ^ symbol --- (?-s).*\R(?=TEST$) 25.6 s Option 'Match Case' unchecked (?-s).+\R(?=TEST$) 23.1 s Option 'Match Case' unchecked (?-s).*+\R(?=TEST$) 21.5 s ( Atomic ) Option 'Match Case' unchecked (?-s).++\R(?=TEST$) 19.2 s ( Atomic ) Option 'Match Case' unchecked (?-s).++\r\n(?=TEST$) 18.1 s ( Atomic ) Option 'Match Case' unchecked (?-s).++\r\n(?=TEST$) 17.25 s ( Atomic ) Option 'Match Case' checked (?-is).++\r\n(?=TEST$) 237 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked --- Without the (?-s)^ part --- .*\R(?=TEST$) 25.1 s Option 'Match Case' unchecked .+\R(?=TEST$) 22.8 s Option 'Match Case' unchecked .*+\R(?=TEST$) 21.2 s ( Atomic ) Option 'Match Case' unchecked .++\R(?=TEST$) 18.9 s ( Atomic ) Option 'Match Case' unchecked .++\r\n(?=TEST$) 17.8 s ( Atomic ) Option 'Match Case' unchecked .++\r\n(?=TEST$) 17 s ( Atomic ) Option 'Match Case' checked (?-i).++\r\n(?=TEST$) 236 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked --- Using the @Terry-R solution --- (?-s).\R(?=TEST$) 77 s ( ?! ) Option 'Match Case' unchecked (?-s).\r\n(?=TEST$) 71 s ( ?! ) Option 'Match Case' unchecked (?-s).{1}+\R(?=TEST$) 93 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked (?-s).{1}+\r\n(?=TEST$) 84 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked .{1}+\R(?=TEST$) 88 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked .{1}+\r\n(?=TEST$) 79 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked --- After CONCATENATION of the line BEFORE the line TEST with the line TEST --- First, the regex \R(?=TEST$) is replaced with NOTHING ( 57 s ) => 142,908,950 bytes for 3,030,301 lines. Then : TEST$ 20.8 s Option 'Match Case' unchecked TEST$ 13.6 s Option 'Match Case' checked (?-i)TEST$ 13.8 s Option 'Match Case' unchecked Last, the regex TEST$ is replaced with \r\n$0 ( 65 s ) -------------------------- HP Win 10 - N++ 8.7.6 ----------------------------------------------------------------------------------------------------- (?-s)^.*\R(?=TEST$) 2,3 s Option 'Match Case' unchecked (?-s)^.*\R(?=TEST$) 2 s Option 'Match Case' checked ( The test in my **previous** post ) (?-s)^.+\R(?=TEST$) 2.3 s Option 'Match Case' unchecked (?-s)^.*+\R(?=TEST$) 1.9 s ( Atomic ) Option 'Match Case' unchecked (?-s)^.++\R(?=TEST$) 1.97 s ( Atomic ) Option 'Match Case' unchecked (?-s)^.++\r\n(?=TEST$) 1.86 s ( Atomic ) Option 'Match Case' unchecked (?-s)^.++\r\n(?=TEST$) 1.5 s ( Atomic ) Option 'Match Case' checked (?-is)^.++\r\n(?=TEST$) 5.8 s ( Atomic ) ( ! ) Option 'Match Case' unchecked --- --- Without the ^ symbol --- (?-s).*\R(?=TEST$) 2.7 s Option 'Match Case' unchecked (?-s).+\R(?=TEST$) 2.3 s Option 'Match Case' unchecked (?-s).*+\R(?=TEST$) 2.3 s ( Atomic ) Option 'Match Case' unchecked (?-s).++\R(?=TEST$) 1.9 s ( Atomic ) Option 'Match Case' unchecked (?-s).++\r\n(?=TEST$) 1.8 s ( Atomic ) Option 'Match Case' unchecked (?-s).++\r\n(?=TEST$) 1.45 s ( Atomic ) Option 'Match Case' checked (?-is).++\r\n(?=TEST$) 23.9 s ( Atomic ) ( !? ) Option 'Match Case' unchecked --- --- Without the (?-s)^ part --- .*\R(?=TEST$) 2.7 s Option 'Match Case' unchecked .+\R(?=TEST$) 2.23 s Option 'Match Case' unchecked .*+\R(?=TEST$) 2.23 s ( Atomic ) Option 'Match Case' unchecked .++\R(?=TEST$) 1.8 s ( Atomic ) Option 'Match Case' unchecked .++\r\n(?=TEST$) 1.7 s ( Atomic ) Option 'Match Case' unchecked .++\r\n(?=TEST$) 1.38 s ( Atomic ) Option 'Match Case' checked (?-i).++\r\n(?=TEST$) 24 s ( Atomic ) ( ?! ) Option 'Match Case' unchecked --- Using the @Terry-R solution --- (?-s).\R(?=TEST$) 6.35 s ( ! ) Option 'Match Case' unchecked (?-s).\r\n(?=TEST$) 8.6 s ( ! ) Option 'Match Case' unchecked (?-s).{1}+\R(?=TEST$) 8.2 s ( Atomic ) ( ! ) Option 'Match Case' unchecked (?-s).{1}+\r\n(?=TEST$) 10.6 s ( Atomic ) ( ! ) Option 'Match Case' unchecked .{1}+\R(?=TEST$) 7.5 s ( Atomic ) ( ! ) Option 'Match Case' unchecked .{1}+\r\n(?=TEST$) 9.75 s ( Atomic ) ( ! ) Option 'Match Case' unchecked --- After CONCATENATION of the line BEFORE the line TEST with the line TEST --- First, the regex \R(?=TEST$) is replaced with NOTHING ( 26.2 s ) => 142,908,950 bytes for 3,030,301 lines. Then : TEST$ 2.3 s Option 'Match Case' unchecked TEST$ 0.95 s Option 'Match Case' checked (?-i)TEST$ 1 s Option 'Match Case' unchecked Last, the regex TEST$ is replaced with \r\n$0 ( 25.3 s )
Conclusion :
So, given the rules above, the best syntaxes seem to be, on my new
Windos 10machine :-
The regex
.++\r\n(?=TEST$)in1.38second, with theMatch Caseoption checked. -
The regex
TEST$in0.95second, AFTER an initial contatenation of the line before the lineTESTwith the lineTEST.
Best Regards,
guy038
-
