Support specifying XPT character variable length via "xpt.length" attribute when writing files

[d2yh]

Currently, when haven writes XPT files, the length of character variables is determined by the maximum length of the strings in the column. I would like to request a feature where, instead of always using the maximum observed length, haven checks for a specified attribute (e.g., "xpt.length") on the R object or column. If this attribute is present, its value should be used as the length upper limit for the character variable in the XPT file. This would provide users with more control over the output and ensure compatibility with systems that expect specific column widths.

Thank you for considering this feature!

[gowerc]

Just as an FYI, if you were looking for a workaround you can simply just pad out your character values e.g.

dat <- data.frame(
    string = stringr::str_pad("hello", width = 30, side = "right")
)
haven::write_xpt(dat, "data/string.xpt")

As in with the above code haven will assign a length of 30 for that character column.

[d2yh]

Hi gowerc,

Thank you for providing your technical solution!

However, I'd like to note that xpt files can indeed exist as "empty files"—that is, they may contain no data rows, but they are still able to carry and describe metadata.
Because of this, I still hope that haven can provide a feature to specify the text length for character variables in the output metadata, even when the file has no data rows.

Thanks again for your input and help!

[gorcha]

Hi @d2yh,

This should already be supported via the width attribute - if the width attribute is not set it defaults to the max string length. The following should work:

dat <- data.frame(
    string = structure("hello", width = 30)
)
haven::write_xpt(dat, "string.xpt")