I have the weirdest problem and my extremely basic knowledge of SQL must be terribly wrong, but I cannot make sense of the behaviour illustrated below.
I have this file test.csv
id,field
A,0
B,1
C,2
D,"0"
E,"1"
F,"2"
G,
H,""
I," "
And this test code:
#! /usr/bin/perl
use strict;
use warnings;
use DBI;
use Devel::VersionDump qw(dump_versions);
my $dbh = DBI->connect ("dbi:CSV:");
$dbh->{RaiseError} = 1;
$dbh->{TraceLevel} = 0;
my $i = 0;
foreach my $cond ("TRUE",
"field <> 0 AND field <> 1",
"field = 0 OR field = 1",
"NOT (field = 0 OR field = 1)",
"NOT field = 0 OR field = 1",
"field <> 0",
"NOT field <> 0",
) {
print "Condition #" . $i++ . " is $cond:\n";
my $sth = $dbh->prepare("SELECT * FROM test.csv WHERE $cond");
$sth->execute();
$sth->dump_results();
};
print "\n\n";
dump_versions();
When run, this is the output:
Condition #0 is TRUE:
'A', '0'
'B', '1'
'C', '2'
'D', '0'
'E', '1'
'F', '2'
'G', ''
'H', ''
'I', ' '
9 rows
Condition #1 is field <> 0 AND field <> 1:
'C', '2'
'F', '2'
'G', ''
'H', ''
'I', ' '
5 rows
Condition #2 is field = 0 OR field = 1:
'A', '0'
'B', '1'
'D', '0'
'E', '1'
4 rows
Condition #3 is NOT (field = 0 OR field = 1):
'A', '0'
'B', '1'
'D', '0'
'E', '1'
4 rows
Condition #4 is NOT field = 0 OR field = 1:
'B', '1'
'C', '2'
'E', '1'
'F', '2'
'G', ''
'H', ''
'I', ' '
7 rows
Condition #5 is field <> 0:
'B', '1'
'C', '2'
'E', '1'
'F', '2'
'G', ''
'H', ''
'I', ' '
7 rows
Condition #6 is NOT field <> 0:
'A', '0'
'D', '0'
2 rows
Perl version: v5.16.3 on MSWin32 (C:\Program Files\Perl64\bin\perl.exe)
ActivePerl::Config - Unknown
ActiveState::Path - 1.01
AutoLoader - 5.73
C:::Program Files::Perl64::site::lib::sitecustomize.pl - Unknown
Carp - 1.26
Class::Struct - 0.63
Clone - 0.34
Config - Unknown
Config_git.pl - Unknown
Config_heavy.pl - Unknown
Cwd - 3.40
DBD::CSV - 0.41
DBD::File - 0.42
DBI - 1.631
DBI::DBD::SqlEngine - 0.06
DBI::SQL::Nano - 1.015544
Data::Dumper - 2.139
Devel::VersionDump - 0.02
DynaLoader - 1.14
Encode - 2.49
Encode::Alias - 2.16
Encode::Config - 2.05
Encode::Encoding - 2.05
Errno - 1.15
Exporter - 5.67
Exporter::Heavy - 5.67
Fcntl - 1.11
File::Basename - 2.84
File::Spec - 3.40
File::Spec::Unix - 3.40
File::Spec::Win32 - 3.40
File::stat - 1.05
IO - 1.25_06
IO::Dir - 1.1
IO::File - 1.16
IO::Handle - 1.33
IO::Seekable - 1.1
List::Util - 1.27
Math::BigFloat - 1.997
Math::BigInt - 1.998
Math::BigInt::Calc - 1.997
Math::Complex - 1.59
Math::Trig - 1.23
Params::Util - 1.07
SQL::Dialects::AnyData - 1.405
SQL::Dialects::Role - 1.405
SQL::Eval - 1.405
SQL::Parser - 1.405
SQL::Statement - 1.405
SQL::Statement::Function - 1.405
SQL::Statement::Functions - 1.405
SQL::Statement::Operation - 1.405
SQL::Statement::Placeholder - 1.405
SQL::Statement::RAM - 1.405
SQL::Statement::Term - 1.405
SQL::Statement::TermFactory - 1.405
SQL::Statement::Util - 1.405
Scalar::Util - 1.27
SelectSaver - 1.02
Symbol - 1.07
Text::CSV_XS - 1.07
Tie::Hash - 1.04
Time::HiRes - 1.9725
Win32 - 0.47
XSLoader - 0.16
base - 2.18
bytes - 1.04
constant - 1.25
integer - 1.00
overload - 1.18
overloading - 0.02
sort - 2.01
strict - 1.07
unicore::Heavy.pl - Unknown
unicore::lib::Perl::Word.pl - Unknown
unicore::lib::Perl::_PerlIDS.pl - Unknown
utf8 - 1.09
utf8_heavy.pl - Unknown
vars - 1.02
warnings - 1.13
warnings::register - 1.02
Condition #0 shows the complete dataset and is fine.
Condition #1 is just some compound condition and works fine.
Condition #2 is the opposite condition (basic logic rules used to invert it), and works fine too.
Yet, condition #3 should be the opposite of #2 and thus equal to #1, but the result is the same as #2: I cannot make any sense of this.
Condition #4 shows that, omitting the parentheses, NOT does work fine, but of course this query is different from any of the previous ones.
Conditions #5 and #6 show a situation where NOT acts exactly as one would expect.
So, why NOT on a compound condition acts as if the NOT were not specified at all?!
By the way, I read this scary post Perl DBD::CSV - SQL Syntax - "AND" clause is not working properly and added Devel::VersionDump to check whether I have a similar issue but it seems to me that all relevant packages are the newest available. Hence, I really have no clue about this.