read_sav fails on zipped files when using col_select

[jokroese]

Problem

When working with large .sav files, I want to both zip them (to reduce their storage space) and read them in with col_select (to reduce their space in memory). Unfortunately, when doing both (reading a zipped file with col_select), haven::read_sav breaks.

Reprex

data <- tibble::tibble(x = 1:5, y = letters[1:5])

haven::write_sav(data, path="test.sav")

zip("test.zip", "test.sav")

# this works
test_works <- haven::read_sav(
  unz("test.zip", "test.sav")
)

# but this fails
test_fails <- haven::read_sav(
  unz("test.zip", "test.sav"),
  col_select = c(x)
)
# Error in isOpen(con) : invalid connection

Workaround

There's a simple workaround:

unz("test.zip", "test.sav")
haven::read_sav(
  "test.sav",
  col_select = c(x)
)
file.remove("test.sav")

but I do this often enough that I'd appreciate it working natively in haven without a function implementing the above workaround.

[gorcha]

Hi @jokroese,

Since we use readr data sources you can read from the zip directly using:

test_fails <- haven::read_sav(
  "test.zip",
  col_select = c(x)
)

I'm also having a look at a potential workaround for using unz() (this affects any complex connection objects used with col_select) but using the zip directly will work fine for simple zipped files.

Alternatively with SAV files you can also use compress = "zsav" to use native SPSS compression.