read_sas fails to import sas character missing values as r missing values

[rogerjdeangelis]

Greate Package.

Issue with: Converting imported sas character missings to r missing values using haven package

CONTENTS

 1 ISSUE (R DOES NOT RCOGNIZE SAS CHARACTER MISSING VALUES)
 2 CORRECTION (using dplyr language)

github (for a beeter view)
https://tinyurl.com/5cxkxfz2
https://github.com/rogerjdeangelis/utl-converting-imported-sas-character-missings-to-r-missing-values-using-haven-package

R version 4.4.1 (2024-06-14 ucrt) -- "Race for Your Life"
Copyright (C) 2024 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64

I tried to paste the issue in this comment box, but it mangled the text. If interested see github

INPUT
=====

CHR ISMISSING

a 0
1
b 0
1

options validvarname=upcase;
libname sd1 "d:/sd1";
data sd1.have ;
input chr $1.;
ismissing=missing(chr);
cards4;
a
.
b
.
;;;;
run;quit;

 PROCESS
 =======

1 ISSUE

%utl_rbeginx;
parmcards4;
library(haven)
library(dplyr)
have<-read_sas("d:/sd1/have.sas7bdat")
have$R_MISSING<-is.na(have$CHR)
print(have)
str(have)
;;;;
%utl_rendx;

---

2 CORRECTION

%utl_rbeginx;
parmcards4;
library(haven)
library(dplyr)
have<-read_sas("d:/sd1/have.sas7bdat")
%>% mutate(across(where(is.character)
,~na_if(.,"")))
have$R_MISSING<-is.na(have$CHR)
print(have)
str(have)
;;;;
%utl_rendx;

    OUTPUT
    =======

R DOES NOT HONOR SAS MISSING

    SAS

CHR ISMISSING R_MISSING
========= =========
1 "a" 0 FALSE
2 "" 1 FALSE WRONG
3 "b" 0 FALSE
4 "" 1 FALSE WRONG

---

   SAS

CHR ISMISSING R_MISSING
========= =========
a 0 FALSE
1 TRUE CORRECTED
b 0 FALSE
1 TRUE CORRECTED

RUNNING IN SAS GUI

LOG (DOSE RUN IN R GUI_

have<-read_sas("d:/sd1/have.sas7bdat")
have$R_MISSING<-is.na(have$CHR)
print(have)

A tibble: 4 × 3

CHR ISMISSING R_MISSING

1 "a" 0 FALSE
2 "" 1 FALSE
3 "b" 0 FALSE
4 "" 1 FALSE

str(have)
tibble [4 × 3] (S3: tbl_df/tbl/data.frame)
$ CHR : chr [1:4] "a" "" "b" ""
$ ISMISSING: num [1:4] 0 1 0 1
$ R_MISSING: logi [1:4] FALSE FALSE FALSE FALSE

[gowerc]

The code blocks in your post appear to be badly formatted which makes it hard to read / tell exactly what you are trying to highlight.

That being said I think you may have a slight miss understanding about SAS character values. SAS doesn't have an explicit missing value for characters and instead represents them as a "". From R's perspective it is impossible to tell what is a blank string vs a missing string so instead just uses a blank string. You can test this with something like the following:

data char_miss ;
input chr $1. chr2 $1.;
ismissing=missing(chr);
chr2 = "";
if chr = chr2;
cards4;
a
.
b
.
;;;;
run;quit;

To be honest its worse than this, because of how SAS stores strings internally SAS makes no difference between trailing spaces on strings. That is to say as far as SAS is concerned " " = " " or "dog " = "dog "

data char_miss ;
	length chr $1. ;
	chr = "A";
	if "dog " = "dog       ";
run;

In short I think haven is working here as intended and this can likely be closed.

[gorcha]

Thanks @gowerc for the explanation!

Yes, confirming haven is running as intended. Empty strings can be manually converted to NA in R if this behaviour is desired.