[Groff] A trap with 'tbl'!

[Ted Harding]

Greetings All!
I'm writing to flag up a little trap that one might
fall into when using 'tbl'.

Background: I'm setting up a table and I want each
row to have a row number to the left of the table.

So I have defined a "left margin number" string \*[lmn]
to do that for each row. My first setup was to define
\*[lmn] so as to compute the row-number incrementally,
then move to the left, print it, and then move back:

.nr lmno -0 1
.ds lmn \Z'\v'-1.5m'\h'-\w'\\n+[lmno]\0\0\0\0\0'u'\\n[lmno]'
.TS H
expand tab(#);
ce cw(2.0i) ce.
Name (in Capitals)#Address#Signature
_
.TH
.T&
|l |l |l|.
\*[lmn]
_
\*[lmn]
_
\*[lmn]
_
\*[lmn]
_
\*[lmn]
_
\*[lmn]
_
.TE

(i.e. with 6 rows). I expected that on being encountered
in the first row, the string \*[lmn] namely

  \Z'\v'-1.5m'\h'-\w'\n+[lmno]\0\0\0\0\0'u'\n[lmno]'

(after interpretation of the definition) would increment
\n[lmno] from 0 to 1, move to the left, print \n[lmno] = 1,
and then move back to the first field of that row. Then,
in the next row, \n[lmno] would be incremented from 1 to 2,
printed on the left, and so on.

What actually happened (I'll illustrate here using the
above 6-row table), was that the first number was 6, not 1.

If you have more rows, say 10, then the number of the
first row will be 10. And so on.

This happens because somehow \n+[lmno] gets interpreted
for each row in the table before it is printed for the
first row. Hence (in this example) it has been incremented
from 0 to 6 at the time it is printed for the first row.
Thereafter it is incremented by 1 for each successive
row, so prints as 6, 7, 8, 9, 10, 11 for this example.

The work-round is to change the initial definition of
\n[lmno] to (in this example):

  .nr lmno -6 1

or (in general)

  .nr lmno -N 1

where N is the number of rows in the table. Then \n[lmno]
is at 1 when the first row is about to be printed.

I have yet to work out how this comes about. When I look at
the output of 'pic' for the above input, I first see:

.nr lmno 0 1
.ds lmn \Z'\v'-1.5m'\h'-\w'\n+[lmno]\0\0\0\0\0'u'\n[lmno]'
.TS H
[nasty pic-made troff code]
.lf 11
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
.lf 13
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
.lf 15
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
.lf 17
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
.lf 19
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
.lf 21
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
.lf 23
.nr 3w0 \n[3w0]>?\w\[tbl]\*[lmn]\[tbl]
[more pic-made troff code]
.TE


So the definition of \*[lmn] has been sent to troff as
expected, and then there are 6 lines (for the 6 lines
of the table), and that is it as far as occurrences
of \*[lmn] are concerned. So it would seem that troff
has only seen \*[lmn] 6 times!

Anyway, that's it, up for discussion if anyone is so
inclined, and in any case as a warning of unexpected
behaviour!

Notes: The intended output of this is a table of rows of
blank boxes, for people to write their names, addresses
and signatures in, with a row-number to the left of each row.

Also: I know I could have done this by making a table
with an extra numeric column of un-ruled boxes on the left,
and put the entry \n+[lmno] into that column in each row.
But I thougt I'd try to be clever ...

Best wishes to all,
Ted.

--------------------------------------------------------------------
E-Mail: (Ted Harding) <address@hidden>
Fax-to-email: +44 (0)870 094 0861
Date: 29-Jul-11                                       Time: 22:25:56
------------------------------ XFMail ------------------------------