@MichaelChirico
The original reported issue with grepl and the use here are distinct cases.
There is not much difference in testing for newlines between grep and grepl.
What was slow is the anchored search at the beginning of the string - clearly this should run much faster and should be fixed in base R

input <- paste(rep("1,2,3,4.567,some text field", 1e7),collapse="\n")
system.time(length(grep("\n", input)))
##   user  system elapsed 
##  0.000   0.000   0.001 
system.time((grepl("\n", input)))
##   user  system elapsed 
##      0       0       0 
system.time((grepl("^http://", input)))
##   user  system elapsed 
##  2.683   0.000   2.685 

True, though checking file.exists is faster than checking for newlines then checking file.exists.

It would seem grepl("\n", ., fixed = TRUE) || grepl("\r", ., fixed = TRUE) is faster still, noting we have already checked that length(.) == 1 a few lines earlier.

@HughParsonage
the point here is, if input is the actual raw data - and therefore potentially very large - then file.exists is really slow and eating up most of execution time of fread.

see my comment in #2531 with the benchmark of file.exists

If on the other hand input is a file name, then we can safely assume that it is a relatively short string (not megabytes) and having the length(grep('\\n|\\r', input)) check before file.exists does not add a real penalty for small inputs.

then length(.)==1 check you refer to is checking something else - whether the input is a vector with length > 1, which fread does not accept.

Codecov Report

Merging #2630 into master will not change coverage.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master    #2630   +/-   ##
=======================================
  Coverage   93.03%   93.03%           
=======================================
  Files          61       61           
  Lines       12115    12115           
=======================================
  Hits        11271    11271           
  Misses        844      844

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8ab3028...1fb6c48. Read the comment docs.