Avoid index name clashing with UNIQUE KEY

When we create a foreign key constraint for MySQL we also create an
index, if there is not already an index beginning with the field in
question. UNIQUEness constaint are implemented by indices, and so we
should also skip adding an index if one already exists.

To avoid changing existing fully-working schemas, this change only skips
index creation if one with the same name already exists. This case
currently fails with a duplicate key error in MySQL, eg.,

    ERROR 1061 (42000): Duplicate key name 'foo'

Author: davel

lib/SQL/Translator/Producer/MySQL.pm

@@ -460,6 +460,16 @@ sub create_table
     #
     my @constraint_defs;
     my @constraints = $table->get_constraints;
+
+    # Mark fields which are first in a UNIQUE constraint as indexed if the
+    # constraint name is the same as the field. This is to prevent a duplicate
+    # index being created to support foreign keys.
+    for my $c ( @constraints ) {
+        if ($c->type eq UNIQUE && $c->name eq ($c->fields())[0]) {
+            $indexed_fields{ ($c->fields())[0] } = 1;
+        }
+    }
+
     for my $c ( @constraints ) {
         my $constr = create_constraint($c, $options);
         push @constraint_defs, $constr if($constr);

t/38-mysql-producer.t

@@ -280,7 +280,6 @@ my @stmts = (
   `id` integer NOT NULL,
   `foo` integer NULL,
   `foo2` integer NULL,
-  INDEX (`foo`),
   UNIQUE `foo` (`foo`, `foo2`),
   PRIMARY KEY (`id`),
   CONSTRAINT `foo` FOREIGN KEY (`foo`) REFERENCES `thing` (`id`)