Re: [S] Error message, S+4 rel3

Prof Brian Ripley (
Mon, 9 Feb 1998 23:09:25 +0000 (GMT) wrote:
> I'd like to relay the experience we've had at Koch Industries related to
> Splus's Version 4.0 Release 3. It may help with others who have Badih's
> question. I'm still discussing this issue with Mathsoft to try to get a
> better understanding of the issues.
> I believe that Chuck Taylor, Mathsoft, already answered this question a
> week or two ago for those running vanilla Splus. His comments regarded
> the __sum4.txt and __sum4i.txt files in the local _Data directory.
> What we've learned is that old Version 3.3 libraries, which do not
> contain the __sum4.txt or __sum4i.txt files but only the __sum.txt file,
> could no longer be accessed by Version 4.0 users who applied the Release
> 3 update. Our fix was to invoke Splus from those directory locations so
> that Splus would create the __sum4.txt and __sum4i.txt files. At this
> point, both our V3.3 and V4.0 users were able to access the libraries.
> Here's our setup. We have a Koch-developed library called trading on a
> shared network drive, which I'll call t:, that everyone accesses via the
> following .First function.
> ".First"<- function() {
> assign("lib.loc", "t:\\spluswin\\library", where =3D 0)
> library(trading)
> }
> Most of our folks already had this .First file in a local _data
> directory on their PC. When they applied the Release 3 update, they got
> the error message
> Error: Cannot read summary data from summary filename
> t:\spluswin\library\trading\_Data/__sum4.txt
> This error didn't go away until I created the
> t:\spluswin\library\_Data\__sum4.txt and __sum4i.txt files on our shared
> network drive. I should mention that our libraries don't have any
> compiled C or Fortran code so we don't face those difficult issues. I
> know that Brian Ripley has done a nice job of making several statlib
> libraries accessible by both V3.3 and V4.0 users in such instances,
> though I don't understand much about how he accomplished this nice feat
> for us. But thanks much, Brian.
I seem to have less problem with this than other people: certainly
one problem is with NT sytems and permissions when the update is done.
On W95 my home directory got fixed during the upgrade: on NT it did not.

My strategy in building libraries is effectively to dump in __BIG
format, and copy the __BIG, __BIGIN and __db3.1 files to a new _Data
directory. I then

- start splus 3.3 in the parent of that directory, which creates __sum.txt
- start Splus4 somewhere else, use library(), then search to find the
number (usually 13) and run make.DB.summary(13) (or whatever). This
creates __sum4*.

Some caveats: (a) I don't have an object browser open by default.
(b) I usually use sqpe.exe for the Splus4 step.

I have only ever seen problems when a __sum4.txt existed. I've just
checked: I can use libraries without one quite happily.

You really don't want to see the scripts I use to build the V&R2 libraries!
They use all of perl, bash and batch files.

> Let me add that I was reluctant to use one Mathsoft proposed solution,
> data.dump and data.restore of the relevant library, because I didn't
> believe this would allow us easy access to all of our help files that
> tended to be created over time, as were the objects in the library. I
> suppose, in time, we could get these all re-sequenced; but that sounded
> like a lot of work. I think when Mathsoft proposed this solution to me
> they didn't really understand the finer points of my questions --
> probably a poor explanation to them on my part.

Actually, I think data.dump and data.restore is quite a good idea.
I take the point about help files (assuming these are nroff files
in _Data\_Help) but believe it is much better to build Windows help
files. I've crafted tools (with Mathsoft DAPD help) to build these
( and to use them ( on 3.3, part of 4.0).
So it is a long time since I worried about re-mapped names.

Brian D. Ripley,        
Professor of Applied Statistics,
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595
This message was distributed by  To unsubscribe
send e-mail to with the BODY of the
message:  unsubscribe s-news