Added comment to describe the proposed fix in the test

diff --git a/t/98rows_prefetch.t b/t/98rows_prefetch.t
index 8619c13..10bc2a9 100644 (file)
--- a/t/98rows_prefetch.t
+++ b/t/98rows_prefetch.t
@@ -10,6 +10,7 @@ use DBICTest;
 plan $@ ? (skip_all => 'needs DBD::SQLite for testing') : (tests => 2);
 
 my $schema = DBICTest->init_schema();
+$schema->storage->debug(1);
 my $no_prefetch = $schema->resultset('Artist')->search(
        undef,
        { rows => 3 }
@@ -28,6 +29,12 @@ my $use_prefetch_count = 0;
 
 is($no_prefetch->count, $use_prefetch->count, '$no_prefetch->count == $use_prefetch->count');
 
+# The fix is to, when using prefetch, take the query and put it into a subquery
+# joined to the tables we're prefetching from. This might result in the same
+# table being joined once in the main subquery and once in the main query. This
+# may actually resolve other, unknown edgecase bugs. It is also the right way
+# to do prefetching. Optimizations can come later.
+
 TODO: {
        local $TODO = "This is a difficult bug to fix, workaround is not to use prefetch with rows";
        $no_prefetch_count++  while $no_prefetch->next;