Home / Questions / 502309
Accepted
D. Albarracín
Asked: 2022-05-10 07:06:36 +0800 CST 2022-05-10 07:06:36 +0800 CST

Show the data type of columns from a csv

  • 772

I am finishing an exercise, and I need to know what type each column of a CSV is (Integer, decimal, date...) through bash.

This is the part that displays the data type, but it doesn't work properly. Also it only shows if it is charor number. Is there any other way to simplify it?

 cat dataset_cars.csv | sed -rn '1p;2s/([^,][[:alpha:]]+[^,])+/(char)/g;2s/([^,][[:digit:]]+[^,])+/(num)/gp'

When I run it, it returns the following, it looks like it doesn't do it really well (it adds zeros, and parentheses where it shouldn't):

,price,brand,model,year,title_status,mileage,color,vin,lot,state,country,condition
0,(num),(char),(char),(num),(char),(num)0,(char), (char(num),(num),(char),(char),(num)char)

And the output I expect should be something like this:

Test data:

Columna1, Columna2, Columna3
prueba, 123, 123.89

Departure:

Columna1, Columna2, Columna3
texto, numerico, decimal
linux

1 Answers

  1. Best Answer

    Despite what I put in the comments, you could use this regex to sed:

    1p;2s/([0-9]*[a-zA-Z]+[0-9]*)+/(string)/g;2s/([[:digit:]]+\.[[:digit:]]+)/(decimal)/g;2s/([[:digit:]]+)/(int)/g;2s/[[:blank:]]//gp

    Let's spread it out a bit so it doesn't look so stuck and cryptic:

    $ sed -rn '1p
2s/([0-9]*[a-zA-Z]+[0-9]*)+/(string)/g
2s/([[:digit:]]+\.[[:digit:]]+)/(decimal)/g
2s/([[:digit:]]+)/(int)/g
2s/[[:blank:]]//gp' archivo.txt

    Where I assume the file archivo.txthas this content type:

    Columna1, Columna2, Columna3, col4, col5, col6,col7
row1, 123, 123.89,r,2.3, 0, n1um
row2, 456, 321.89,s,0.1, 1, n2um

    And the output gives something like:

    Columna1,Columna2,Columna3,col4,col5,col6,col7
(string),(int),(decimal),(string),(decimal),(int),(string)

    In short, the first thing we do with the command is print the first line. Then I go looking for patterns in the second line (which according to my understanding and without checking) where each value is shown to be a string, decimal, or integer.

    In the end, I just remove whitespace from our result that isn't needed for type checking; only for the presentation of the result.

    Clearly this is a weak way of approaching it, since we do not take into account separators or characters that define a text, such as single or double quotes. But as long as it serves, you can take it.

    Perhaps my answer could be of more help to you with awk: scrpiting - show the data type of each column

    • 2