in reply to How do I test for a NULL string?

By saying "But none of those are a one-to-one match with 'NULL'", you both deny a simple answer and leave some information off the table that would be helpful to those coming at this from an SQL perspective.

The simple answer, which is probably "right" for much of the audience is to explain undef and how you test for it. For someone fetching a row of fields out of a database, where some of the field can be NULL, this answer is good enough.

Where this answer goes awry is in SQL comparison semantics. In SQL, you test for the presense of NULL using an "IS NULL" clause, because NULL is never equal to NULL. A frequent source of SQL newbie errors is preparing   SELECT a FROM t WHERE b = ? and then binding NULL (or undef) to the placeholder. No go. One needs to write   SELECT a FROM t WHERE b IS NULL Granted, this is a SQL problem and not a Perl problem, but it's lurking there waiting to bite the uninformed. In my opinion, it's worth mentioning whenever any asks about how to do NULL in Perl.

Replies are listed 'Best First'.
Re: Re: How do I test for a NULL string?
by merlyn (Sage) on Feb 12, 2002 at 19:44 UTC
    The simple answer, which is probably "right" for much of the audience is to explain undef and how you test for it.
    You show your SQL bias as much as djantzen shows his C bias!

    There is no "right" answer for "much of the audience", except to ask what is meant. And that's my original point. My litany of possible responses is meant to describe how many "right" answers there could be possibly, and no answer is more right than any of the others, except determined by context!

    -- Randal L. Schwartz, Perl hacker

[reply]

      1. The phrase 'NULL string' has a well-known and precise meaning in C ('""').

      2. Whatever meaning this phrase has in SQL is derived from it's meaning in C. It does not have the well-known and precise meaning in the former as in the latter and the question 'How do I test for a SQL NULL string?' is better rephrased 'How do I test if a char/varchar field declared 'NULL' contains a C-style NULL string ('""')?'.

      3. Employing the eminent philosopher of language Donald Davidson, I introduce "the principle of charity", according to which a listener wishing to interpret a speaker accurately "must maximize the self-consistency attribut(ed) to him, on pain of not understanding him" (From "Truth and Meaning").

      While the term 'NULL' is obviously part of SQL, and it is not impossible to define the phrase 'NULL string' within the semantics of SQL, it does not follow that it is equally likely that the questioner was in fact importing those semantics rather than the semantics of C.

      Similarly, although the romance languages share cognates derived from Latin, take libre for example, when I speak Spanish I am not also speaking French by virtue of the shared term. In such a situation the listener must determine from the context which language I am speaking. If I make the utterance "Creo en habla libre, y cerveza gratis", it is wholly incorrect to believe that I am speaking French, but poorly, because an interpretation of the utterance that attributes self-consistency to me is available. To conclude that I am speaking French on the basis of the appearance of "libre" is to attribute to me a great deal of nonsense surrounding that one term, and this is not sound interpretation.

      Therefore, because there is a well-known and precise meaning available to the phrase 'NULL string', and because I believe that the principle of charity will yield in general better interpretation, to conclude that the speaker refers to it's meaning in C is a reasonable conclusion, indeed, more reasonable than any other, including SQL.

      Now according to merlyn's clarifications, his point is that the listener must be sensitive to different meanings possibly at work in a question. This is fine, however I would suggest that an orthogonal pressure here ought to be the principle of charity. It is better to risk thinking too highly of your interlocutor than to assume immediately that his utterances are largely nonsense. If the point here is not to give a real answer to the question 'How do I test for a NULL string', for which there is a sensible answer in Perl, but to give an exhortation to be careful about importing meanings carelessly across language boundaries, then I suggest the question be rephrased in such a way that the principle of charity not demand a particular interpretation. The rather more innucuous "How do I test for a string that is null?" would make the point better.

[reply]
        You continue to miss the point here.
        • You cannot control how they ask it.
        • You cannot presume what context they are using.
        • You cannot even presume that they are perfect in constructing the question even if you could know the context.
        You can theorize all you want. In practice, I've heard "NULL string" mean all of the things I've listed, whether that was right, wrong, or even possible! And learning to deal with that ambiguity makes me a better answerer.

        -- Randal L. Schwartz, Perl hacker

[reply]
Re: Re: How do I test for a NULL string?
by demerphq (Chancellor) on Feb 13, 2002 at 10:32 UTC
    ..is to explain undef and how you test for it.

    Problem being 

    my ($x,$y);
print "Same!" if $x==$y;
__END__
Same!

    [download]
    Since it doesnt follow the properties of the SQL null i'm not real sure why explaining undef will help any more than explaining an empty string.

    Yves / DeMerphq
    --
    When to use Prototypes?

[reply]
[d/l]
[select]

In Section Meditations