I don't agree that the user-assigned coverage ID provides a better link than
the internal coverage number.
I've done some work with m.in.e00, both using it and coding for it. I recall
that when I imported a point coverage I had to use the -i option, but when I
imported a line coverage the -i option did not work correctly. My
recollection could be faulty.
More recently I modified m.in.e00 to correctly read polygon coverages (that
modification has not yet been merged back to CVS) and found that the e00
format does not use the coverage-ID in the polygon topology (PAL) table. I
couldn't find a way to use the coverage-ID to link the data; it appeared
necessary to use the coverage-#.
I thought that it might be best to remove the -i option entirely.
I can think of only one problem that might come up when indexing with
coverages numbers. You cannot import data from two different e00 files and
use the category numbers to compare objects between the two imported data
sets. There is no way to guarantee that the category number for an object
from one file will be the same category number for the same object from a
different file.
Roger Miller
On Thu, 19 Jun 2003 16:34:35 -0400 (EDT), Michael Ash wrote
Could I ask for a bit more clarification on this matter? I
have had some difficulty with .e00 files and the associated
dig_cats/ files that are created.
When I have chosen the "-i" option of m.in.e00 and then
copied the resulting dig/, dig_att/, and dig_plus/ files to
correspond to the various dig_cats/ files, I have mismatches
between areas on the map and the associated dig_cats labels.
The problem does not occur when I omit the "-i" option of
m.in.e00. So my conclusion had been to omit the "-i"
option.
However, Michel's post suggests that "-i" is the appropriate
option because coverage-# can be shuffled by subsequent
processing while coverage-ID is fixed. I would be grateful
for clarification.
Best regards,
Michael
> Keith J. Forbes wrote:
> > FLAG: Set the following flag?
> > Link attributes by coverage-ID not by coverage-#?
> > Question: What exactly does this option mean?
>
> Each feature in e00 file (and in Arc/Info) have an internal
> identifier (name_of_layer-#) to link attributes and geometry.
> It is however a bad practice to use this number as an external
> link because the internal numbering may change if you do
> some operation (cleaning after erasing some objects, etc.).
> # is computed automaticaly and you cannot modify it.
> There is another identifier (name_of-layer-ID) that is
> entirely controled by the user, and can therefore be used
> to link your coverage to an external database, etc...
>
> In practice for m.in.e00 :
> - the is one attribute entry per graphic object (line, area, point)
> so you should use the normal behaviour of the program.
> - in some case, you want to consider all objects having the
> same ID as one object in Grass, or the # links are broken
> (yes, sometimes you can get really bad datasets !), so you
> may want to try the ID number.
>
> Honestly, its also an historical feature : the first release
> used (incorrectly) ID for the link...