Support binary compression in sas7bdat about readstat

[xiaodaigh]

Compression support is definitely needed! I have to stick to the parso library in Java until this is supported.

[xiaodaigh]

anyone has example files,

That is usually the problem!

I try to keep track of it here

https://github.com/xiaodaigh/sas7bdat-resources

[sclewis23]

If anyone has example files, please try them with this new code branch:

https://github.com/WizardMac/ReadStat/tree/sas-binary-compression

I'll create a sample tomorrow.

[evanmiller]

@ofajardo Thanks for testing – I will wait a few days for results from other files and then merge if everything looks okay.

[evanmiller]

@sclewis23 Is this the correct data?

"IDnumber","week1","week16","AverageLoss"
"2477",195.000000,163.000000,32.000000
"2431",220.000000,198.000000,22.000000
"2456",173.000000,155.000000,18.000000
"2412",135.000000,116.000000,19.000000

[ajdamico]

hi, not sure if this is the right place for a minimal reproducible example since tidyverse/haven#31 was closed? the latest version of haven still fails on compressed data.. thanks

# devtools::install_github('biostatmatt/sas7bdat.parso')
library(haven)
library(sas7bdat.parso)


tf1 <- tempfile()

download.file( "http://www.census.gov/housing/extract_files/data%20extracts/cpsasec14/hhld.sas7bdat" , tf1 , mode = 'wb' )

# fails
x1 <- read_sas( tf1 )

# works
y1 <- read.sas7bdat.parso( tf1 )

[hadley]

@evanmiller is this on the schedule?

[evanmiller]

No schedule for this -- note that many compression issues were misdiagnosed as binary compression rather than bugs in the character decompressor. Seems like 90%+ compressed files in the wild are character compressed.

[evanmiller]

Note that the example file provided by @ajdamico was fixed in 69d5751 and 8c0463a.

[reikoch]

If a truly binary compressed SAS dataset is needed, you may use [https://github.com/reikoch/testfiles/blob/master/binary.sas7bdat]. haven 1.0 fails with "ReadStat: Error parsing page 0, bytes 8192-16383".

[sclewis23]

Is this encoding error related to compression ?
Unsupported character set code: 204.
tidyverse/haven#482

[evanmiller]

@sclewis23 No - the error is related to the file's character encoding.

If you know which encoding was used to create the file, I can try to add support.

[sclewis23]

@evanmiller - the encoding is set to "any"
outencoding=any

[sclewis23]

@evanmiller
here is a sample SAS program, creates one with UTF-8 and "any" :

#SAS CODE:
# Data Weight2; 
# input IDnumber $ week1 week16; 
# AverageLoss=week1-week16; 
# datalines; 
# 2477 195 163
# 2431 220 198
# 2456 173 155
# 2412 135 116
# ;
# libname outlib '~' outencoding='UTF-8';
# data outlib.Weight_utf8;
# Set Weight2;
# Run;
# libname out_any '~' outencoding='any';
# data out_any.Weight2;
# Set Weight2;
# Run;

[sclewis23]

@evanmiller
I can send example files if that helps?

[evanmiller]

If anyone has example files, please try them with this new code branch:

https://github.com/WizardMac/ReadStat/tree/sas-binary-compression

[ofajardo]

tested OK in pyreadstat with the attached sample file I generated in SAS like this:

data SAMPLES.sample_bincompressed(compress=binary);
set SAMPLES.sample;
run;

The file is stored permanently in the pyreadstat repo in the test_data/basic/sample_bincompressed.sas7bdat, for now in the sasbin_dev branch.

sample_bincompressed.sas7bdat.zip

[sclewis23]

Here is a very small sample file with binary compression(4 rows).

weigth2.zip

[sclewis23]

Looks good:

[reikoch]

Looks good on my two testfiles dates_binary.sas7bdat and dates_longname_binary.sas7bdat in https://github.com/reikoch/testfiles - congratulations!

[evanmiller]

@reikoch Thanks for letting me know. The reports are all positive so I'll get this merged into dev later today.

[evanmiller]

Merged into master and included in 1.1.4 - closing