That was my expectation as well. I am surprised (pleasantly) that it doesn't produce an error and it appears to give accurate results, but I can't find any documentation on it. I'm sure it is there somewhere in help. I found out in testing that with sql, if you don't convert it using isnull(), it can give erroneous results. For instance when the value of a bit field allows nulls:

(ISNULL(myfield,0)!=1)

instead of:

myfield!=1

but in VFP9 it doesn't appear to be necessary to convert the null value before checking it.


>Looks like it works without it too for TRUE case, though I would assume to give an error instead.
>

>ll=.t.
>ll=null
>=messagebox(iif(ll,"True","False"))
>
>ll=.f.
>ll=null
>=messagebox(iif(not ll,"False","True"))
>

>
>In other words, I would stay with NVL to give exact result.
>
>
>>Is there anytime that a field which allows nulls would generate an error on the code:
>>
>>

>>IF myfield.myvalue  && is the value .t.?
>>

>>
>>The field value can be .T., .F., or null
>>
>>It is never generating an error, but I have always had the code:
>>
>>

>>IF (NVL(myfield.myvalue,.F.))
>>

>>
>>and am wondering if the NVL() is guaranteed to not be necesssary? I have tested the speed difference
>>and even with over a million records the NVL() speed difference appears negligible. At onetime the
>>NVL() was added to all fields (all types char, date, log, etc) as a safety check more or less.
>>
>>TIA,
>>Tracy