issue with whitespace parsing in read_table

[phil-grayson]

Hi,

I've just completed a Genome Wide Association run of PLINK! | v1.07 | 10/Aug/2009

One of the output files there is space delimited and after some googling, I found that read_table is the tidyverse solution. I'm running 1.3.1 of readr in R 4.0.0. I've tried the following (the head -100 of the file can be found attached here for a test).

assoc_readr <- read_table("readr_test.txt", col_names = TRUE)
this one parses the first 3 columns correctly and then lumps the last 7

assoc_base <- read.table("readr_test.txt", sep = "" , header = T)
this one does it right. 10 columns parsed on whitespace. the whole file is just shy of a gig, so I had wanted to use tidyverse to load it in (love that progress bar!)

(https://github.com/tidyverse/readr/files/5012278/readr_test.txt)

I was able to get it all sorted out with base R, but wanted to make an issue so that it can be fixed if it is a larger problem. cheers!

[jimhester]

readr::read_table() uses the same width for all columns.

The function analogous to utils::read.table() is readr::read_table2(), which breaks on whitespace.

This is mentioned in ?read_table

‘read_table2()’ is like ‘read.table()’, it allows any number of
whitespace characters between columns, and the lines can be of
different lengths.

read_table()’ is more strict, each line must be the same length,
and each field is in the same position in every line. It first
finds empty columns and then parses like a fixed width file.

[postylem]

Just a side note for clarification:

readr::read_table() uses the same width for all columns.

I don't think that's true, if I'm interpreting you right. With read_table(), the columns don't have to all be the same width as each other, but any one column is the same width for all rows. In contrast, read_table2() will break on any whitespace, so columns don't have to be aligned.

For example, if myfile.txt contains:

NAME                         X  Y    Z
very-long-entry              1  2    3
short                        1  1  2.3
fi fi fo fum                 3 23  101
something                    3 -1  s 0 

then read_table('myfile.txt') gives

# A tibble: 4 x 4
  NAME                X     Y Z    
  <chr>           <dbl> <dbl> <chr>
1 very-long-entry     1     2 3    
2 short               1     1 2.3  
3 fi fi fo fum        3    23 101  
4 something           3    -1 s 0  

while read_table2('myfile.txt') gives the following (with warnings about parsing failures)

# A tibble: 4 x 4
  NAME            X     Y     Z    
  <chr>           <chr> <chr> <chr>
1 very-long-entry 1     2     3    
2 short           1     1     2.3  
3 fi              fi    fo    fum  
4 something       3     -1    s    

[jennybc]

Things have changed since this issue was current.

https://readr.tidyverse.org/news/index.html#deprecated-or-superseded-functions-and-features-2-0-0

In readr 2.0.0, released 2021-07-20 (after this issue thread concluded):

read_table2() has been renamed to read_table(), as most users expect read_table() to work like utils::read.table(). If you want the previous strict behavior of the read_table() you can use read_fwf() with fwf_empty() directly (#717).

[postylem]

Thanks. I got here from a search, and was confused for a bit, so thought I'd make note for others who might land here too. Thanks for clarifying.