Re: [Gretl-devel] Fwd: Re: slight inconsistency with strstr truth value
On Mon, 12 Mar 2018, Allin Cottrell wrote:
Ah, good, getting rid of strings-as-booleans shouldn't be too
painful then.
I've found lots of tests involving strstr in StrucBreak, cointARDL, PTconf,
ADMBP, johansensmall, a_eff, margeff_el, FEP, waldTest and lagreg. But in
almost all cases the usage is one or other of
if strstr(s1, s2) != ""
if strlen(strstr(s1, s2))
(which can now be accomplished in git via instring(s1, s2)). As you say the
FEP GUI seems to be the only exception.
So I'm now thinking, is there much point in delaying the switch?
With a very minor update to FEP all our packages would be OK, and
the string-as-boolean thing was never documented; seems to me only
those following a rather arcane discussion 1.5 years ago would even
be tempted in that direction, and apparently the temptation has not
been strong.
If there's strong objection I'll back off, but I'm inclined to
remove string-as-boolean from gretl 2018a (with a message in the
backward-incompatibility log, of course).
Allin