MathGroup Archive 1998

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Corrupted NB

  • To: mathgroup at smc.vnet.net
  • Subject: [mg14518] Re: Corrupted NB
  • From: gwinn at ma.ultranet.com (Joe Gwinn)
  • Date: Thu, 29 Oct 1998 04:33:16 -0500
  • Organization: Gwinn Instruments
  • References: <70dd69$234@smc.vnet.net> <70rkmi$4u5$9@dragonfly.wolfram.com>
  • Sender: owner-wri-mathgroup at wolfram.com

I have been bitten by this too.

There is a simple way Wolfram could at least make Mathematica aware that
a prefs file had become corrupted (by other than Mathematica?) -- put a
simple checksum on it.  This would allow Mathematica to know that the
file was damaged, and tell the user, rather than to become weird.

By the way, there is a simple trick I sometimes use on apps that too
often eat their own prefs files -- manually make a backup copy, and/or
manually write-lock the active prefs file, so the app cannot write to
its own prefs file.  Most apps will tolerate this, although I haven't
tried it with Mathematica.  It is a bit of a nuisance, as one must
unlock the prefs file to change the prefs, but has been effective in
the past.

A variant of this would be to require Mathematica to get specific user
permission to write to the prefs file: a last-chance "commit" button. 
This, combined with a checksum, should greatly reduce the level of
trouble from pref file damage, even as the bulk of Mathematica evolves
and changes. 

Joe Gwinn

In article <70rkmi$4u5$9 at dragonfly.wolfram.com>, Garrett Tim Sos
<gts at mindspring.com> wrote:

> Maybe this idea should also apply to the preferences files that still
> (version 3.0.1.1x) become corrupted (on the Mac, atleast) so often that
> I no longer try to change from the factory default preferences.
> 
> I logged a call to WRI on a different issue and asked about  the
> preferences file corruption problem and got the answer that it was "the
> computer's fault" implying that they did not feel it needed to be
> fixed.
> 
> Maybe he was just having a bad day but sometimes it is hard  to like
> WRI.  I have been using Mathemetica since  Version 1.1.  It is a
> wonderfully complex creation but 3.0's habit of destroying user INPUT
> (preferances, notebooks and strange corruption of cells) is dreadful.
> 
> Tim
> 
> 
> Joe Gwinn wrote:
> > 
> > I have also had this problem, but I have never succeeded in fixing a
> > damaged notebook.  The problem has always happened when Mathematica
> > crashed while saving the notebook  Now I periodically make a no-output
> > backup copy called "<whatever>.nb short" so I don't lose everything
> > when Mathematica crashes.
> > 
> > This kind of destroy-file-on-save behaviour was very common in early
> > (mainframe and mini) computer software, and subsequently in early PC
> > and Mac software, but is now mostly a thing of the bad old days.  The
> > standard solution is to do file saves in such a way that there is never
> > a window of vulnerability, where a computer crash (for whatever reason)
> > could leave one with only a corrupted file.
> > 
> > The method?  Simple.  One 1970s variant:  Open a temporary file and
> > write the new version of the document into it.  If successful, rename
> > the old file with a temporary name, and then rename the just-written
> > temporary file (containing the new document) to be the updated file.
> > Delete the old file with the temporary name.  No matter when the crash,
> > there is at least one good file surviving, perhaps with a temporary
> > name.  Temporary files are not deleted on a crash, so this can work.
> > 
> > This does require that one have scratch disk space sufficient to carry
> > an extra copy of the file, but this is generally true, especially in
> > this day of multi-gigabyte disks.  More to the point, this method was
> > used even when disks were tiny, so one can conclude that users would
> > rather have bulletproof file saves than the extra disk space.
> > 
> > Joe Gwinn
> > 
> > In article <70dd69$234 at smc.vnet.net>, "Kevin J. McCann"
> > <kevinmccann at Home.com> wrote:
> > 
> > > I have a NB which when I try to run it tells me that there was a syntax
> > > error.  The popup goes on to say that I can open it as a text file,
> > > etc.  Is there a way to recover from this.  The problem appears to have
> > > been generated when Mathematica crashed.
> > >
> > > Kevin


  • Prev by Date: CDF[MultinormalDistribution[...]] only for n < 3?
  • Next by Date: Re: ParametricPlot
  • Previous by thread: Re: Re: Corrupted NB
  • Next by thread: Plot, Cursor and Spelling Errors questions