Thanks to you both. I'm happy to have it confirmed as a no-op.

Integer-boolean conversion, not needed here though perhaps in an earlier
version of J, or indeed APL. That makes sense.

Cheers.

On Wed, Feb 3, 2016 at 10:35 AM, Roger Hui <[email protected]>
wrote:

> I don't think it's that significant.  In 1=?s$2 the 1= is a no-op. It's
> there possibly because:
>
> - Muscle memory.  In APL, in 1-origin, the 1= is needed to convert integers
> to booleans.
> - There's a lot of cut-and-paste in making the test scripts.  Could have
> been cut and pasted from somewhere else.
> - Phrase of the moon.
>
> If anyone knows, I should know (I wrote the test script), and I don't
> know/remember.
>
>
>
> On Tue, Feb 2, 2016 at 2:22 PM, Matthew Baulch <[email protected]>
> wrote:
>
> > I've been reading the tests in the J engine source. Have learned quite a
> > bit while doing so! Anyhow, I've become perplexed by one particular line.
> >
> > In the 'Boolean' section of the tests for 'x i. y' (test/gi.ijs), a
> random
> > table is constructed with
> >
> > a=:1=?10 5$2
> >
> >
> > My question is, why is a=:?10 5$2 not sufficient. Can someone tell be if
> > these are equivalent, or if not, why?
> >
> >
> > I should note that 1=?70$3 is used in the same file, and it's clear to me
> > that '1=' affects the probability distribution for rolls over a domain
> > larger than two, so it serves a purpose.
> >
> >
> > Sorry if source would have been a better forum for this post. I'm not
> sure
> > if this is an error in the test, or simply a deficiency in my
> understanding
> > of J.
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to