cavern 0.95pre5

[Olly Betts]

In message <39C12375.B0016BDF@dybal.eti.br>, Leandro Dybal Bertoni writes:
>It runs OK here, but I would like to ask about two things:
>
>- What does this warning mean?
>
>1999_02/cafe_3.svx:11: warning: Reentering an existing prefix level is
>deprecated (previously entered at 1996_01/cafe.svx:9)
>
>The "offending" line in 1999_02/cafe_3.svx is:
>
>*begin \c

It means you've also got a "*begin \c" in 1996_01/cafe.svx, line 9.

Especially in larger projects, this sort of arrangement is probably a bad
idea, if not totally accidental.  Adding this warning revealed at least 2
places in the CUCC Austria dataset where the same survey had been included
twice.

>- What does the new info in the .err file mean?

It means I was working on the gross error detector in Austria, but forgot to
turn it off again.

>traverse \a2.58->\a2.52 e=(0.01, -0.06, -0.10) mag=0.11 suspect:
>\a2.58->\a2.52 G: 0.05 (-6.00deg -> -6.98deg)

This means that the traverse's error (which has magnitude 11cm) is greater
than can reasonably (95% confidence) be attributed to random error.  But if
you change the gradient on the leg a2.58 to a2.52 from -6 to -6.98 that
reduces the magnitude of the traverse's error to 5cm which can reasonably be
attributed to random error.

If changes to the length (L) or bearing (B) of any legs in the traverse also
resulted in a residual error which could reasonably be attributed to random
error, these would also be listed, along with the magnitudes of the residual
errors.  In this case it's a one leg traverse, so it's not suprising that we
only get one candidate.

It works fairly well, though in multiple looped surveys the gross error
distorts the end points which the algorithm is trying to match to and so you
don't get useful suggestions for the traverses with the larger errors (which
is where gross error detection would be most useful).  The candidate
traverse really needs disconnecting while the analysis is performed, which
is hard to do in the current code.

I'll switch it off again for now.  I might add an option to turn it on
before 1.0.

Cheers,
Olly