4 views (last 30 days)
Show older comments
Christos Antonakopoulos on 16 Nov 2015
-
-
Link
Direct link to this question
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings
Commented: Jenny Smith on 19 Jul 2018
Accepted Answer: Guillaume
Open in MATLAB Online
My file is about 72 MB, almost 850000 rows and on average 7 columns, so some times the number of columns changes. Data is mostly comprised of strings so i used the:
http://www.mathworks.com/matlabcentral/fileexchange/23573-csvimport
as
name= 'etch.csv';
[C1, C2, C3, C4, C5, C6, C7] = csvimport(name, 'columns', [1:7], 'noHeader', true, 'delimiter', ';' );
(i am interested only in the 7 columns even there were cases with more data) This works perfectly for small data sets. For my case it took me almost 30 minutes or even more. Any idea for something better? Thank you
PS My data type is:
1: Device Name,Category,Date,Time,Source,Message,Condition,Name,Act
2: string1,string2,mm/dd/yyyy,hh:mm:ss.sss,string,string,string,1 or 0
.....
850000: and it goes on as line 2
last column most of the times has no data but does not interest me
2 Comments Show NoneHide None
Show NoneHide None
Mohammad Abouali on 16 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323558
have you tried readtable?
Christos Antonakopoulos on 17 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323618
it does not work for my case since the number of delimiters is not always the same
Sign in to comment.
Sign in to answer this question.
Accepted Answer
Guillaume on 17 Nov 2015
Open in MATLAB Online
No matter what, you're bound by the reading speed of matlab. Probably the fastest way to read the file is to rea it all once with fileread. You can then split the lines with strsplit. It is then a choice of applying either of textscan, strsplit or regexp on each line. You would have to see which is faster.
Here is how I would do it using regexp:
filecontent = fileread('etch.csv');
filelines = strsplit(filecontent, {'\r', '\n'}); %split at line ending. Copes with linux and windows termination
fields = regexp(filelines, '^([^;]*);([^;]*);([^;]*);([^;]*);([^;]*);([^;]*);([^;]*);', 'tokens', 'once'); %only keep the first seven fields
fields = vertcat(fields{:})
The above takes about 3 seconds on my machine to read 85000 rows (only 8 MB of text though).
One thing it hasn't done is parse the date. This is fairly trivial to do with datetime if needed and takes no time at all.
4 Comments Show 2 older commentsHide 2 older comments
Show 2 older commentsHide 2 older comments
Christos Antonakopoulos on 17 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323656
Thank you, your code exactly worked for my 850000 lines i needed about 70 seconds, when before i needed 30 minutes. I will check the other commands to see which is the quickest.
Guillaume on 17 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323688
Open in MATLAB Online
Oh, I was out by a factor of 10 on my test data. That explains the difference in size.
From my testing, the longest operation is the strsplit into individual lines. I've just realised that this operation is not actually needed and that the exact same regular expression I wrote can be used on the whole file. You just need to change one option of the regex:
filecontent = fileread('etch.csv');
fields = regexp(filecontent, '^([^;]*);([^;]*);([^;]*);([^;]*);([^;]*);([^;]*);([^;]*);', 'tokens', 'lineanchors');
fields = vertcat(fields{:})
On my machine, it halves the processed time (5 seconds vs 11 seconds for 850000 lines, you need a better computer!).
Also note that the vertcat is also an expensive operation. If you're happy to access your data as a cell array (lines) of cell arrays (tokens), then you can dispense with it.
Finally, note that if a line has 7 or less fields, the regex won't match. That can be worked out by modifying the regex at the expense of more processing time.
Christos Antonakopoulos on 18 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323862
I see, yes you are right my time was also reduced, but still i need a better pc. Thank you again
Jenny Smith on 19 Jul 2018
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_591065
Hello, I am trying to follow this thread and I'm reading through the regex documentation... I don't understand what you are doing with this expression with [^;]* I have a very similar problem, my text is separated by commas and I have seven columns, and I am trying to understand how to use this function similarly.
Sign in to comment.
More Answers (1)
dpb on 16 Nov 2015
Open in MATLAB Online
Can't do anything w/o at least a sample of the data file with whatever warts there are as far as missing fields, but why not go to the root i/o routines directly? For larger fields, "as near to the metal as you can get" is bound to be the ploy.
fmt='%s %s %2d/%2d/%2d %2d:%2d:%2d %s %s %s %*[^\n]';
d = textscan(fid,fmt,'delimiter',',','headerlines',1);
The result above will be a cell array of 7xN; if you do want the various variables then try same format string with textread instead.
Note there's a new %d formatting string with latest release to parse dates on input directly; I don't have past R2012b so return the m/d/y and h/m/s as numerics above. If you do want to retain the strings instead and do the conversion later (or perhaps don't need them any other way) it should be obvious where to replace the formatting to do so.
2 Comments Show NoneHide None
Show NoneHide None
dpb on 16 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323525
ADDENDUM OBTW, it might turn out to be faster to use a looping construct and read a smaller subset of the file each pass rather than the whole thing at once...with textscan you can pick up from previous read automagically; textread in this regards always closes the file so it would have to reopen it every time with an updated 'headerlines' argument; probably a losing proposition.
I don't know if this would help or not; you'd just have to 'spearmint to see if less memory requirements per read operation would outperform the alternate.
Christos Antonakopoulos on 17 Nov 2015
Direct link to this comment
https://matlabcentral.mathworks.com/matlabcentral/answers/255103-import-too-large-csv-data-file-with-strings#comment_323620
Edited: Stephen23 on 17 Nov 2015
Open in MATLAB Online
Device Name;Category;Date;Time;Source;Message;Condition Name;Act;Ack;Ena
CCT AC800 PEC Local;Event;08/26/2010;16:47:09.9550;PEC_MSG_25_10;SerialCommFault;Active;1;1;1
CCT AC800 PEC Local;Trip;08/26/2010;16:46:50.2530;PEC_MSG_1_08;LineUndervoltage;Active;1;1;1
CCT AC800 PEC Local;Trip;08/26/2010;16:46:50.2530;PEC_MSG_1_11;LineUnderfrequency;Active;1;1;1
CCT AC800 PEC Local;Trip;08/26/2010;16:47:09.9550;PEC_MSG_26_10;WaterPressure Fault;Active;1;0;1
That are exactly the first 5 lines, i am not interested on the last 3 columns though. As i said there are cases, in which my rows have less than 10 or more than 10 columns, that is why with csvimport function i had my problem solved since those cases were solved through padding or truncation.
Sign in to comment.
Sign in to answer this question.
See Also
Categories
MATLABData Import and AnalysisLarge Files and Big Data
Find more on Large Files and Big Data in Help Center and File Exchange
Tags
- import csv files
- strings
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
An Error Occurred
Unable to complete the action because of changes made to the page. Reload the page to see its updated state.
Select a Web Site
Choose a web site to get translated content where available and see local events and offers. Based on your location, we recommend that you select: .
You can also select a web site from the following list
Americas
- América Latina (Español)
- Canada (English)
- United States (English)
- Belgium (English)
- Denmark (English)
- Deutschland (Deutsch)
- España (Español)
- Finland (English)
- France (Français)
- Ireland (English)
- Italia (Italiano)
- Luxembourg (English)
- Netherlands (English)
- Norway (English)
- Österreich (Deutsch)
- Portugal (English)
- Sweden (English)
- Switzerland
- Deutsch
- English
- Français
- United Kingdom(English)
Asia Pacific
- Australia (English)
- India (English)
- New Zealand (English)
- 中国
- 日本Japanese (日本語)
- 한국Korean (한국어)
Contact your local office