test1=: 0:`1:`2: @. (3&|)
For J9.01 and earlier,
test1 i. 6 is a rank error because execution of (3&|) doesn't produce
an atom.
So you guys must be using J9.02. Thanks for helping out!
Starting in 9.02, this doesn't automatically produce a rank error.
Instead, the result of (3&| i. 6/7) is inspected. It is seen to produce
a result with the same shape as (i. 6): that is, there is one
result-atom for each atom of (y). This means that the implied rank of
the verbs in (0:`1:`2:) is 0: each verb is assumed to produce a result
for each 0-cell (i. e. atom) of input.
[Note that this is a requirement on the gerund-verb. The gerund is not
executed with rank 0: that would be inefficient. Rather, the gerund
must behave properly when given an argument of rank > 0.]
[if the selector-verb were (3 | +/"1) and the argument were (i. 3 6),
the selector would produce a result with shape (,3) and the JE would
conclude that the implied rank of the gerund-verbs is 1.]
The gerunds are then executed as called for by the result of (3&|). How
many times are they executed? That's none of your business - the JE
might execute them one at a time, or it might group them together. When
you execute (test1 i. 6), the JE executes each gerund on an atom, and
you get no error.
But when you execute (test1 i. 7), the JE decides differently. It
collects the inputs to the first gerund (0 3 6) and passes them in a
single call to the verb (0:). Remember, the verb is charged with giving
a result on each atom. But this verb doesn't. It gives a single atomic
result, which is a rank error because it is not the 3 results expected.
How to fix it? Make each gerund follow the rules, and give a result on
each atom. A simple way, inefficient in general, is to make sure the
rank of each gerund is the implied rank of the verb:
test1a=: 0"0`(1"0)`(2"0) @. (3&|)
test1a i. 7
0 1 2 0 1 2 0
The gerund-verb can have a higher rank if it is well-behaved:
test1b=: 0"0`(3&|)`(3&|) @. (3&|)
test1b i. 7
0 1 2 0 1 2 0
This is what I was trying to say in the NuVoc writeup. If you would
reread it and tell me what's unclear, or just fix the Wiki yourself, I'd
be obliged.
Henry Rich
On 6/7/2020 1:18 PM, Devon McCormick wrote:
I see the same anomaly for the "i.7" argument on J 9.02.03 under Windows 10:
test1=: 0:`1:`2: @. (3&|)
test1 &.> i.&.> >:i.6
+-+---+-----+-------+---------+-----------+
|0|0 1|0 1 2|0 1 2 0|0 1 2 0 1|0 1 2 0 1 2|
+-+---+-----+-------+---------+-----------+
test1 i.7
|rank error: test1
|test1[0]
dbr 1 [ dbr 0
test1 i.100
|rank error: test1
|test1[0]
But defining "test1" with scalar rank does not exhibit this behavior:
test1=: 0:`1:`2: @. (3&|)"0
test1&.>i.&.>7+i.3
+-------------+---------------+-----------------+
|0 1 2 0 1 2 0|0 1 2 0 1 2 0 1|0 1 2 0 1 2 0 1 2|
+-------------+---------------+-----------------+
On Sun, Jun 7, 2020 at 10:39 AM Hauke Rehr <[email protected]> wrote:
test1 i.6
doesn’t work for me either
but both
test1"0 i.6
and
test1"0 i.7
do work as expected
Am 07.06.20 um 16:35 schrieb çağlar girit:
Hi,
Although I have read NuVoc entries for Agenda and Constant Function, I
still don't understand why for "test1 =. 0: ` 1: ` 2: @. (3&|)", "test1
i.
6" works whereas "test2 i.7" gives a rank error. Am I missing some
subtlety of implied rank with Agenda?
test1 =. 0: ` 1: ` 2: @. (3&|)
test1 i.6
0 1 2 0 1 2
test1 i.7
|rank error: test1
| test1 i.7
Thanks!
--
----------------------
mail written using NEO
neo-layout.org
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
--
This email has been checked for viruses by AVG.
https://www.avg.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
