Corrupting files on windows Readr 2.0

[l-gorman]

Hi all!

I recently upgraded to readr version 2.0. When I did this, I ran in to some problems with older scripts. I worked out, the problem was that if I read a csv using readr::read_csv and then rewrote the object to a file using readr::write_csv, the file would become corrupted in some way.

In particualr, when writing the file, I would get the error:

Cannot open the file for writing

When I went to try and reopen the file in Excel, it would open without a problem, but when trying to modify and save the file, I would receive the error:

Can not be accessed. The file may be corrupted located on a server that is not responding

This is an example of code that would produce this type of error:

# The process was simply
df<-data.frame(read_csv("filename.csv",na=c('n/a','<NA>','NA')))

write_csv(df,"filename.csv")

I am using windows 10 (64-bit) and R version 4.1.0.

In the end I switched back to version 1.4 and everything works again.

Let me know if you need any more details :)

(By the way this is not a complaint at all, I love this package, just thought I would point this out in case it helpful!)

[jimhester]

With lazy reading a file handle remains open until the whole file has been read. On Windows if any process has an open file handle to a file it is locked and nothing else can write to it.

Use lazy=FALSE when reading the file, and then no file handle will remain open and what you are trying to do will work.

[l-gorman]

@jimhester Thanks so much! That's really good to know!