Overpermissive parsing of strings

[rgayler]

# BUG - [parse|col]_double not consistent with *_numeric and *_integer
# *_double("information") -> Inf
# whereas *_numeric("information) -> NA and *_integer("information") -> NA

# DESIGN - Having *_[numeric|double|integer]("3D graphics) - > 3
# May be convenient but strikes me as dangerous.
# How about including a parameter "strict = TRUE | FALSE"
# strict = FALSE gives the current behaviour 
# strict = TRUE gives NA if there are any problems
# I would argue for strict = TRUE as the default

library(readr) # V 0.1.1 from CRAN

# make an input text
in_stream <- "information\n3d graphics"
in_vec <- c("information", "3d graphics")

# parse it different ways
xpn <- parse_numeric(in_vec) # 1 problem
xpd <- parse_double(in_vec)  # 2 problems
xpf <- parse_float(in_vec)  # no such function
xpi <- parse_integer(in_vec) # 2 problems

# collect it different ways
xcn <- read_csv(in_stream, col_names = FALSE, col_types = list(col_numeric())) # 1 problem
xcd <- read_csv(in_stream, col_names = FALSE, col_types = list(col_double()))  # 2 problems
xcf <- read_csv(in_stream, col_names = FALSE, col_types = list(col_float()))  # no such function
xci <- read_csv(in_stream, col_names = FALSE, col_types = list(col_integer())) # 2 problems

xpn
xcn
problems(xcn)

xpd
xcd
problems(xcd)

xpi
xci
problems(xci)

[hadley]

Minimal reprex

parse_numeric("3d")
parse_double("3d")
parse_integer("3d")

I believe the behaviour of parse_numeric() is correct (and as documented), but parse_double() and parse_integer() should return NA, not 3.