NotaBene Mailing List

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Nasty NB change-invisible behaviour (was: DOS text finder)



This is mostly about NB, but first, a quick report on MINITRUE,
the best of the DOS text-in-files  finder I've tried so far.
Maiben Poirier reported that MINITRUE truncated an NB file during
a search. We've corresponded about how it happened, and neither
of us has been able to make it happen again. I've tested MINITRUE
with small, medium, large and huge (860K) NB test files, loaded
to the gunwhales with formatting, asking the program to search
for text, and to search and replace. I did it repeatedly, always
with the same results - MINITRUE behaved impeccably every time.

I wish I could say the same of NB. I tried the same tests on the
same heavily formatted files, asking NB to replace one string
with another - and NB did something nasty. Round about 64K (give
or take a few hundred) it swallowed a few bytes. Say that I had
the following sentence sprinkled throughout the file:

   This is a test sentence.

- and that I was doing a global change of `test' to `new'. In
most places in the file the result was:

   This is a new sentence.

- except that when the sentence was in the immediate vicinity of of
64K, it got changed to:

   This isnewentence.

When the file was up to 300K long, this byte-gobbling took place
only at 64K. With a file of 800-900K long, it happened at 1 or
2 sites further down the file - but unpredictably. It would
happen here in one run, there in the next; whereas in a heavily
formatted file it always happened at 64K.

In a text file with no formatting, however, byte-gobbling didn't
happen at all, even when the file was 900K. Nor did it in a
heavily formatted one shorter than 62K.

I wondered whether it might be a peculiarity of my machine's
interaction with NB, so I copied the test files to my laptop and
tried again. Same result.

It was on the list that I learned that NB could behave peculiarly
around the 64K mark. The discussion at that time suggested, if I
remember right, that NB's frisky habit of swallowing footnotes
or changing formatting code into low-ASCII characters - smiley
faces etc. - was at its worst around 64K. Ever since then, I've
split my work files when they get close to 60K. If I don't, sure
enough, I have trouble. This test seems to suggest that one
should also be extremely cautious about making global-invisible
changes in heavily formatted NB files longer than about 60K. Old
NB hands probably know this already, but I didn't.

Best wishes

Mary

Mary Bernard
<if10000>


Main Index | Thread Index