Add a reprex, and use readLines() to see the difference

file <- tempfile()
writeLines(c("", "", NA), file)
readLines(file)
#> [1] ""   ""   "NA"
data.table::fread(file, blank.lines.skip = FALSE)
#>        V1
#>    <lgcl>
#> 1:     NA
sessionInfo()
#> R version 4.3.1 (2023-06-16)
#> Platform: x86_64-pc-linux-gnu (64-bit)
#> Running under: Ubuntu 22.04.3 LTS
#>
#> Matrix products: default
#> BLAS/LAPACK: /usr/lib/x86_64-linux-gnu/libmkl_rt.so;  LAPACK version 3.8.0
#>
#> locale:
#>  [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8
#>  [4] LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8
#>  [7] LC_PAPER=C.UTF-8       LC_NAME=C              LC_ADDRESS=C
#> [10] LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C
#>
#> time zone: Asia/Shanghai
#> tzcode source: system (glibc)
#>
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base
#>
#> loaded via a namespace (and not attached):
#>  [1] digest_0.6.33     styler_1.10.1     fastmap_1.1.1     xfun_0.39
#>  [5] magrittr_2.0.3    glue_1.6.2        R.utils_2.12.2    knitr_1.43
#>  [9] htmltools_0.5.5   rmarkdown_2.23    lifecycle_1.0.3   cli_3.6.1
#> [13] R.methodsS3_1.8.2 vctrs_0.6.3       reprex_2.0.2      data.table_1.14.9
#> [17] withr_2.5.0       compiler_4.3.1    R.oo_1.25.0       R.cache_0.16.0
#> [21] purrr_1.0.1       tools_4.3.1       evaluate_0.21     yaml_2.3.7
#> [25] rlang_1.1.1       fs_1.6.3

^{Created on 2023-11-03 with reprex v2.0.2}

We should probably update the documentation that blank lines at the beginning of a file are always skipped until the first non-empty row is encountered. This is also mentioned by the verbose output (shortened it to highlight the important part).

library(data.table)
file2 <- tempfile()
writeLines(c("", "", "a"), file2)
fread(file2, blank.lines.skip = FALSE, verbose=TRUE)
#> [05] Skipping initial rows if needed
#>   Positioned on line 3 starting: <<a>>

It’s not the document, we want to completely turn off the skipping of blank lines

I get that but changing the behavior of blank.lines.skip = FALSE would probably break a lot of existing code so I doubt we will do that.

Possible changes I could imagine would be to introduce another argument or let blank.lines.skip accept other values to indicate you do not want to skip blank lines at the beginning.

Do we see a purpose of an additional argument to turn off the initial skipping of lines? Or is a documentation update fine to close this issue, shouldn't be difficult to implement regardless.

fread has already so many arguments so I wouldn't want to introduce a new argument for that specific case. What do you think blank.lines.skip=NULL for indicating this behavior? Other free value of NA seems off, although I would ultimately tend towards blank.lines.skip="none" or something similar

@ben-schwen I did some initial prototyping of a change and here's what I found:

It seems that to make this behavior work in the most intuitive way, (blank.lines.skip="none" doesn't skip any blank lines, for any number of columns) there's likely a lot more work involved as we have to rework how fread() does skipping blank lines with more than 1 column. Not sure if this one use case is enough to warrant this.

Currently, blank.lines.skip=FALSE is ignored if we have more than 1 column of data (not sure if this is documented):
#6158 (comment)

If we make it so blank.lines.skip="none" only works with 1 column which is consistent with current blank.lines.skip=FALSE, then current changes should be most of the way there, barring some more testing. This would solve the MRE posted with the original issue, but I'm pretty sure OP intends to use this feature with data that has more than 1 column, which wouldn't be covered.

ATP, I feel a documentation change is the way to go. If we want to cover the MRE from the issue, the change is almost there. WDYT?

Tough question. Files with multiple columns and complete blank lines are clearly misspecified, so it's hard to guess what the intended behavior should be.

Does it play nice with complete blank lines, multiple columns and fill=TRUE?

f = tempfile()
input = c("a,b", ",", ",")
writeLines(input, f)
fread(f, blank.lines.skip=FALSE, fill=TRUE)
#     a  b
# 1: NA NA
# 2: NA NA

Seems to work, I think fill is what we needed.

I could initialize fill=TRUE whenever blank.lines.skip="none"?

Edit:
The above does work but only because column headers are specified. If the first line is empty, fread doesn't have a good way of defining columns, which is why the default behavior is to skip blanks until we find a non-empty line so we can define columns. If I bypass the skipping of the first blanks, fread just assumes the file to have one column, see:

f = tempfile()
input = "\n\n\n\n1,3\n\n2,4\n\n"
writeLines(input, f)
fread(f, blank.lines.skip="none", fill=TRUE)
#     V1 
#  1: NA                              
#  2: NA                              
# ---                                 
#  8: NA                              
#  9: NA 

Following up on this issue -- I figured that skipping the initial blank lines isn't quite so trivial because AFAIU fread uses the first non-empty line to determine the columns of the result. Do we think this is still worth pursuing to handle this case?

I want to keep the total line number.

Hmm, it seems fread() is not exactly suitable for this. For example, we will attempt to find free-text fields that include newlines but are otherwise properly quoted & can thus be understood as a "rectangular" table.

If we simply want to count the number of newlines, this SO Q&A look pretty good to me (and adjustments to respect \r-delimited lines are not too hard):

https://stackoverflow.com/a/23456450/3576984

As to desired fread() behavior... still not sure... what's the interaction with header=FALSE for the examples here?

what's the interaction with header=FALSE for the examples here?

f = tempfile()
input = c("a,b", ",", ",")
writeLines(input, f)
fread(f, blank.lines.skip=FALSE, fill=TRUE, header=FALSE)
       V1     V2
   <char> <char>
1:      a      b
2:              
3:              

input = "\n\n\n\n1,3\n\n2,4\n\n"
writeLines(input, f)
fread(f, blank.lines.skip="none", fill=TRUE, header=FALSE)
      V1
   <num>
1:    NA
2:    NA
3:    NA
4:    NA
5:   1.3
6:    NA
7:   2.4
8:    NA
9:    NA

fread(f, blank.lines.skip='none')
      V1    V2
   <int> <int>
1:     2     4