write.csv

Two thoughts:

1) Will the new server that you are setting up with Server 2016 eventually host build and test? If so, this could at least run on that.
2) CreateVirtualDisk and OpenVirtualDisk are C functions that are available in Windows 7 and Server 2008 R2. So these are options that we could use, but it would require creating a small program to drive creation/mounting/deletion of the disk and compiling it at build time.

Nathan

-----Original Message-----
From: Uwe Ligges [mailto:ligges at statistik.tu-dortmund.de]
Sent: Tuesday, July 11, 2017 5:09 AM
To: Nathan Sosnovske <nsosnov at microsoft.com>; Duncan Murdoch <murdoch.duncan at gmail.com>; Jim Hester <james.f.hester at gmail.com>
Cc: r-devel at r-project.org; Lipatz Jean-Luc <jean-luc.lipatz at insee.fr>
Subject: Re: [Rd] write.csv

This is a bit difficult:

R binaries (both for R base and R packages) are still built with Windows Server 2008 and my desktop machine is Windows 7, hence at least currently such a check would not get executed on the machines R core / CRAN use ...

Best,
Uwe

On 11.07.2017 00:59, Nathan Sosnovske via R-devel wrote:

As a follow up to this, Martin Maechler suggested that I write a small script that allows this to be tested on Windows from R. The attached script demonstrates creating and mounting a very small (4 MiB) VHD to a path on the system (in this case c:/smallmount) and then calling write.csv with a large dataframe to the newly mounted path.

The method I attached has two limitations.

1) It will only work if run as administrator. This is a limitation of the OS.
2) It will only work on Windows 8.1/Server 2012R2 or higher, as it uses powershell commands that only exist on those operating systems. I believe it could be made to work on Windows 7, but it would need to be written using the windows C apis at that point.

If a regression test is created for this issue I would be more than happy to integrate this method into that if there is interest and if it would work in the environment where automated builds/tests for windows are run.

-----Original Message-----
From: Nathan Sosnovske
Sent: Tuesday, July 4, 2017 8:02 AM
To: 'Duncan Murdoch' <murdoch.duncan at gmail.com>; Jim Hester
<james.f.hester at gmail.com>
Cc: r-devel at r-project.org; Lipatz Jean-Luc <jean-luc.lipatz at insee.fr>
Subject: RE: [Rd] write.csv

The best way to test on Windows would probably be creating a small virtual hard disk (via CreateVirtualDisk), mounting it, and writing to the mounted location. I believe the drive could even be mounted to an arbitrary location on the filesystem (instead of a drive letter) so that drive letter conflicts don't come into play.

-----Original Message-----
From: R-devel [mailto:r-devel-bounces at r-project.org] On Behalf Of
Duncan Murdoch
Sent: Tuesday, July 4, 2017 7:53 AM
To: Jim Hester <james.f.hester at gmail.com>
Cc: r-devel at r-project.org; Lipatz Jean-Luc <jean-luc.lipatz at insee.fr>
Subject: Re: [Rd] write.csv

On 04/07/2017 10:01 AM, Jim Hester wrote:

On linux at least you can use `/dev/full` [1] to test writing to a full device.

      > echo 'foo' > /dev/full

      bash: echo: write error: No space left on device

Unfortunately, I get a permission denied error if I try to write there from MacOS.  I don't know if Windows has an equivalent.

I've taken a look at the code.  Essentially it comes down to a call to the C function vfprintf, which is supposed to return the number of bytes written, or a negative value for an error. This return value is often not checked; in particular, write.table and friends don't check it.

I'll add code to signal an error if there's a negative value.

I don't think it's feasible to check the number of bytes (formatted text with possible translation to a different encoding could have any number of bytes) if it's positive.  So hopefully all of our file systems will correctly signal an error, and not just report how many bytes were successfully written.

Although that won't be a perfect test for this case where part of the
file is written successfully.

An alternative suggestion for testing this is to create and mount a
loop device [2] with a small file.

[1]:
https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fen.w
i
kipedia.org%2Fwiki%2F%2Fdev%2Ffull&data=02%7C01%7Cnsosnov%40microsoft.
com%7Cb97a7371538b4dbe9a7308d4c2ec5aa0%7C72f988bf86f141af91ab2d7cd011
d
b47%7C1%7C0%7C636347767773809248&sdata=Cb2oduozc2IDCLvXZGG1C4i4hQA7FP
s
5jHmnFYbk7zQ%3D&reserved=0
[2]:
https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fstac
k
overflow.com%2Fa%2F16044420%2F2055486&data=02%7C01%7Cnsosnov%40micros
o
ft.com%7Cb97a7371538b4dbe9a7308d4c2ec5aa0%7C72f988bf86f141af91ab2d7cd
0
11db47%7C1%7C0%7C636347767773809248&sdata=%2BWPfqD0nUS%2F30DUNDqQU79l
R
EJh02ZX0yik9HXiY5kg%3D&reserved=0

Loop devices sound ideal, but seem to be Linux-only (at least with that recipe).

Duncan

On Tue, Jul 4, 2017 at 3:38 PM, Duncan Murdoch <murdoch.duncan at gmail.com> wrote:

On 04/07/2017 8:40 AM, Lipatz Jean-Luc wrote:

I would really like the bug fixed. At least this one, because I
know people in my institute using this function.
I understand your arguments about open source, but I also saw in
this mail list a proposal for a fix for this bug for which there
were no answer from the people who are able to include it in the
distribution. It looks like if there were interesting bugs and the other ones.

Please post a link to that, and I'll look.  Bug reports should be
posted to the bug list.  It's unfortunate that it is currently so
difficult to do so, but if they are only posted here, they are often overlooked.

I don't understand the other arguments : the example was reproduced
with a simple USB key and you cannot state that a disk will
eternally be empty enough, specially when it has several users.

I am not denying that it's a bug, I'm just saying that it is a
difficult one to test automatically (so we probably won't add a
regression test once it's fixed), and it's not one that has been
reported often.  I didn't know there were any reports before yours.

Duncan Murdoch

JLL


-----Message d'origine-----
De : Duncan Murdoch [mailto:murdoch.duncan at gmail.com] Envoy? :
mardi
4 juillet 2017 14:24 ? : Lipatz Jean-Luc; r-devel at r-project.org
Objet : Re: [Rd] write.csv

On 04/07/2017 5:40 AM, Lipatz Jean-Luc wrote:

Hi all,

I am currently studying how to generalize the usage of R in my
statistical institute and I encountered a problem that I cannot
declare on bugzilla (cannot understand why).

Bugzilla was badly abused by spammers last year, so you need to
have your account created manually by one of the admins to post there.
Write to me privately if you'd like me to create an account for you.
(If you want it attached to a different email address, that's
fine.)

Sorry for trying this mailing list but I am really worried about
the problem itself and the possible implications in using R in a
professionnal data production context.

The issue about 'write.csv' is that it just doesn't check if there
is enough space on disk and doesn't report failure to write data.

Example (R 3.4.0 windows 32 bits, but I reproduced the problem
with older versions and under Mac OS/X)

fwrite(as.list(1:1000000),"G:/Test")

Error in fwrite(as.list(1:1e+06), "G:/Test") :
    No space left on device: 'G:/Test'

write.csv(1:1000000,"G:/Test")

I have a big concern here, because it means that you could save
some important data at one point of time and discover a long time
after that you actually lost them.

   > I suppose that the fix is relatively straightforward, but how

can we be sure that there is no another function with the same bad properties?

R is open source.  You could work out the patch for this bug, and
in the process see the pattern of coding that leads to it.  Then
you'll know if other functions use the same buggy pattern.

Is the lesson that you should not use a R function, even from the
core, without having personnally tested it against extreme conditions?

I think the answer to that is yes.  Most people never write such
big files that they fill their disk:  if they did, all sorts of
things would go wrong on their systems.  So this kind of extreme
condition isn't often tested.  It's not easy to test in a platform
independent
way:  R would need to be able to create a volume with a small
capacity.  That's a very system-dependent thing to do.

And wouldn't it be the work of the developpers to do such
elementary tests?

Again, R is open source.  You can and should contribute code (and
therefore become one of the developers) if you are working in
unusual conditions.

R states quite clearly in the welcome message every time it starts:
"R is free software and comes with ABSOLUTELY NO WARRANTY."  This
is essentially the same lack of warranty that you get with
commercial software, though it's stated a lot more clearly.

Duncan Murdoch