MATLAB Answers

0

import large csv file

hello, I have a 22 digit count in the first column of my csv file and all the numbers are important. But Matlab has written the number as E21 and losing significant figures, what should I do? and I have more than one question I just want to get the 1st, 3rd and 5th columns, how can I do?

  댓글 수: 0

로그인 to comment.

답변 수: 3

Walter Roberson 님의 답변 11 May 2019
 채택된 답변

You cannot read in the first field as numeric. There is no support in the built in types for 22 decimal digits of precision.
detectImportOptions on the file and set the variable type for the first field to be string and readtable with that options structure.

  댓글 수: 0

로그인 to comment.



Use the readmatrix() command to import the csv file. The 'Output Type' arugment will let you change the data type and 'Range' argument lets you specify the columuns you want to read.

  댓글 수: 3

The other thing you should know is, eventhough readmatrix() reads all the your digits, the command window might not display them, as its set display to the short format by default. You can change this by using the format command
yes I used the readmatrix () command, but what I try is "Undefined function or variable 'readmatrix'." I got the error, and I thought this code couldn't get the csv files.
Could you please upload the code you're trying and the csv file?

로그인 to comment.


Jeremy Hughes 님의 답변 13 May 2019

The issue with 22 digits of precision in DECIMAL is that MATLAB's default number type is DOUBLE which really can only handle around 17 digits at which point the number system itself cannot distinguish the numbers. The import functions are rounding to the nearest double value.
i.e.
1.000000000000000000000001 is the same as 1.0
If the data are all integers, you can specify int64/uint64.
>> A = readmatrix(filename,'OutputType','int64')
If not integers, than Walter's answer is the best you can do.
Jeremy

  댓글 수: 3

int64 can handle at most 9223372036854775807 which is 19 digits. uint64 gets you up to 18446744073709551615 which at least trickles over 1E20 . Neither one can handle 22 decimal digits.
Ahhh yes, you're right.
This makes me curious about where the data comes from. Maybe 80 bit long double?
https://en.wikipedia.org/wiki/Extended_precision#IEEE_754_extended_precision_formats x64 architecture uses 64 bit significand with no "hidden bit", so you would not get any integer precision beyond what you could get with uint64.

로그인 to comment.



Translated by