Don't get min/max a 2nd time in _chunk_char(). Use where in _chunk_char(). Test char chunking world_city.name with real range stats. Add failing pt-table-checksum char chunk --where test.

lib/TableChunker.pm

@@ -204,6 +204,7 @@ sub find_chunk_columns {
 #   exact       - Use exact chunk_size? Use approximates is not.
 #   tries       - Fetch up to this many rows to find a non-zero value
 #   chunk_range - Make chunk range open (default) or openclosed
+#   where       - WHERE clause.
 #
 # Returns:
 #   Array of WHERE predicates like "`col` >= '10' AND `col` < '20'",
@@ -510,6 +511,9 @@ sub _chunk_numeric {
 #                   <TableChunker::get_range_statistics()>
 #   chunk_size    - requested size of each chunk
 #
+# Optional Arguments:
+#   where - WHERE clause.
+#
 # Returns:
 #   Array of chunker info that <calculate_chunks()> uses to create
 #   chunks, like:
@@ -522,7 +526,7 @@ sub _chunk_numeric {
 #   (end code)
 sub _chunk_char {
    my ( $self, %args ) = @_;
-   my @required_args = qw(dbh db tbl tbl_struct chunk_col rows_in_range chunk_size);
+   my @required_args = qw(dbh db tbl tbl_struct chunk_col min max rows_in_range chunk_size);
    foreach my $arg ( @required_args ) {
       die "I need a $arg argument" unless defined $args{$arg};
    }
@@ -533,15 +537,8 @@ sub _chunk_char {
    my $row;
    my $sql;
 
-   # Get what MySQL says are the min and max column values.
-   # For example, is 'a' or 'A' the min according to MySQL?
-   $sql = "SELECT MIN($chunk_col), MAX($chunk_col) FROM $db_tbl "
-        . "ORDER BY `$chunk_col`";
-   MKDEBUG && _d($dbh, $sql);
-   $row = $dbh->selectrow_arrayref($sql);
-   my ($min_col, $max_col) = ($row->[0], $row->[1]);
-
    # Get the character codes between the min and max column values.
+   my ($min_col, $max_col) = @{args}{qw(min max)};
    $sql = "SELECT ORD(?) AS min_col_ord, ORD(?) AS max_col_ord";
    MKDEBUG && _d($dbh, $sql);
    my $ord_sth = $dbh->prepare($sql);  # avoid quoting issues
@@ -642,7 +639,9 @@ sub _chunk_char {
    # [ant, apple, azur, boy].  We assume data is more evenly distributed
    # than not so we use the minimum number of characters to express a chunk
    # size.
-   $sql = "SELECT MAX(LENGTH($chunk_col)) FROM $db_tbl ORDER BY `$chunk_col`";
+   $sql = "SELECT MAX(LENGTH($chunk_col)) FROM $db_tbl "
+        . ($args{where} ? "WHERE $args{where} " : "") 
+        . "ORDER BY `$chunk_col`";
    MKDEBUG && _d($dbh, $sql);
    $row = $dbh->selectrow_arrayref($sql);
    my $max_col_len = $row->[0];

t/lib/TableChunker.t

@@ -27,7 +27,7 @@ if ( !$dbh ) {
    plan skip_all => 'Cannot connect to sandbox master';
 }
 else {
-   plan tests => 88;
+   plan tests => 90;
 }
 
 $sb->create_dbs($dbh, ['test']);
@@ -1174,21 +1174,27 @@ SKIP: {
 
 $sb->load_file('master', "t/lib/samples/char-chunking/world-city.sql", 'test');
 $t = $p->parse( $du->get_create_table($dbh, $q, 'test', 'world_city') );
-@chunks = $c->calculate_chunks(
+%params = $c->get_range_statistics(
-   tbl_struct    => $t,
-   chunk_col     => 'name',
-   min           => 'A Coruña (La Coruña)',
-   max           => '´s-Hertogenbosch',
-   rows_in_range => 4079,
-   chunk_size    => 500,
    dbh        => $dbh,
    db         => 'test',
    tbl        => 'world_city',
+   chunk_col  => 'name',
+   tbl_struct => $t,
+   chunk_size => '500',
+);
+@chunks = $c->calculate_chunks(
+   dbh           => $dbh,
+   db            => 'test',
+   tbl           => 'world_city',
+   tbl_struct    => $t,
+   chunk_col     => 'name',
+   chunk_size    => 500,
+   %params,
 );
 ok(
    @chunks >= 9,
    "At least 9 char chunks on test.world_city.name"
-);
+) or print STDERR Dumper(\@chunks);
 
 my $n_rows = count_rows("test.world_city", "name", @chunks);
 is(
@@ -1247,14 +1253,48 @@ $t = $p->parse( $du->get_create_table($dbh, $q, 'test', 'checksum_test') );
 is(
    $params{min},
    11,
-   'MIN range stats with --where (bug 821673)'
+   'MIN int range stats with --where (bug 821673)'
 );
 is(
    $params{max},
    15,
-   'MAX range stats with --where (bug 821673)'
+   'MAX int range stats with --where (bug 821673)'
 );
 
+# char chunking
+$sb->load_file('master', "t/pt-table-checksum/samples/where02.sql");
+$t = $p->parse( $du->get_create_table($dbh, $q, 'test', 'checksum_test') );
+%params = $c->get_range_statistics(
+   dbh        => $dbh,
+   db         => 'test',
+   tbl        => 'checksum_test',
+   chunk_col  => 'id',
+   tbl_struct => $t,
+   where      => "date = '2011-03-03'",
+);
+is(
+   $params{min},
+   'Apple',
+   'MIN char range stats with --where (bug 821673)'
+);
+is(
+   $params{max},
+   'raspberry',
+   'MAX char range stats with --where (bug 821673)'
+);
+
+# It's difficult to construct a char chunk test where WHERE will matter.
+#@chunks = $c->calculate_chunks(
+#   dbh           => $dbh,
+#   db            => 'test',
+#   tbl           => 'checksum_test',
+#   tbl_struct    => $t,
+#   chunk_col     => 'id',
+#   chunk_size    => 5,
+#   where         => "date = '2011-03-03'",
+#   %params,
+#);
+
 # #############################################################################
 # Done.
 # #############################################################################

t/pt-table-checksum/basics.t

@@ -24,7 +24,7 @@ if ( !$master_dbh ) {
    plan skip_all => 'Cannot connect to sandbox master';
 }
 else {
-   plan tests => 13;
+   plan tests => 14;
 }
 
 my ($output, $output2);
@@ -134,7 +134,25 @@ ok(
       "t/pt-table-checksum/samples/where01.out",
       trf => "awk '{print \$1 \" \" \$2 \" \" \$3}'",
    ),
-   "--where affects range stats (bug 821673)"
+   "--where affects int range stats (bug 821673)"
+);
+
+# Test it again with a varchar primary key.  The resulting 5 rows are:
+# | Apple     | 2011-03-03 |
+# | lemon     | 2011-03-03 |
+# | lime      | 2011-03-03 |
+# | pineapple | 2011-03-03 |
+# | raspberry | 2011-03-03 |
+$sb->load_file('master', "t/pt-table-checksum/samples/where02.sql");
+ok(
+   no_diff(
+      sub { pt_table_checksum::main(@args, 
+         qw(--no-zero-chunk --chunk-size 5), '--where', "date = '2011-03-03'");
+      },
+      "t/pt-table-checksum/samples/where02.out",
+      trf => "awk '{print \$1 \" \" \$2 \" \" \$3}'",
+   ),
+   "--where affects char range stats (bug 821673)"
 );
 
 # #############################################################################

t/pt-table-checksum/samples/where02.sql

@@ -0,0 +1,26 @@
+drop database if exists test;
+create database test;
+use test;
+
+CREATE TABLE `checksum_test` (
+  `id` varchar(255) NOT NULL,
+  `date` date DEFAULT NULL,
+  PRIMARY KEY (`id`)
+) ENGINE=InnoDB;
+
+INSERT INTO `checksum_test` VALUES 
+      ('Apple',       '2011-03-03'), 
+      ('banana',      '2011-03-01'), 
+      ('orange',      '2011-03-01'), 
+      ('grape',       '2011-03-01'), 
+      ('kiwi',        '2011-03-01'), 
+      ('strawberry',  '2011-03-02'), 
+      ('peach',       '2011-03-02'), 
+      ('mango',       '2011-03-02'), 
+      ('tomato',      '2011-03-02'), 
+      ('nectarine',   '2011-03-02'), 
+      ('pear',        '2011-03-01'), 
+      ('lemon',       '2011-03-03'), 
+      ('lime',        '2011-03-03'), 
+      ('pineapple',   '2011-03-03'), 
+      ('raspberry',   '2011-03-03');