Implement support for MySQL bitstrings.

We migrate bit(xx) to the same PostgreSQL datatype bit(xx) where in Postgres
we can use bitstring as documented at the following URL. In particular the
COPY syntax accepts the notation Xabcd for the values, which is quite nice
when MySQL sends the data to us a a byte vector:

  https://www.postgresql.org/docs/current/datatype-bit.html

Fixes #943.

src/sources/mysql/mysql-cast-rules.lisp

@@ -36,10 +36,16 @@
 	     :target (:type "boolean" :drop-typemod t)
 	     :using pgloader.transforms::tinyint-to-boolean)
 
+    ;; bit(1) is most often used as a boolean too
     (:source (:type "bit" :typemod (= 1 precision))
 	     :target (:type "boolean" :drop-typemod t)
 	     :using pgloader.transforms::bits-to-boolean)
 
+    ;; bit(X) might be flags or another use case for bitstrings
+    (:source (:type "bit")
+	     :target (:type "bit" :drop-typemod nil)
+	     :using pgloader.transforms::bits-to-hex-bitstring)
+
     ;; bigint(20) unsigned (or not, actually) does not fit into PostgreSQL
     ;; bigint (-9223372036854775808 to +9223372036854775807):
     (:source (:type "bigint" :typemod (< 19 precision))

src/utils/transforms.lisp

@@ -82,6 +82,7 @@
 		 time-with-no-separator
 		 tinyint-to-boolean
 		 bits-to-boolean
+                 bits-to-hex-bitstring
 		 int-to-ip
 		 ip-range
 		 convert-mysql-point
@@ -180,6 +181,29 @@
         (fixnum    (if (= 0 bit) "f" "t"))
         (character (if (= 0 (char-code bit)) "f" "t"))))))
 
+(defun bits-to-hex-bitstring (bit-vector-or-string)
+  "Transform bit(XX) from MySQL to bit(XX) in PostgreSQL."
+  (etypecase bit-vector-or-string
+    (null nil)
+    ;; default value as string looks like "b'0'", skip b' and then closing '
+    (string (let ((default bit-vector-or-string)
+                  (size    (length bit-vector-or-string)))
+              (subseq default 2 (+ -1 size))))
+    (array  (let* ((bytes  bit-vector-or-string)
+                   (size   (length bit-vector-or-string))
+                   (digits "0123456789abcdef")
+                   (hexstr
+                    (make-array (+ 1 (* size 2)) :element-type 'character)))
+              ;; use Postgres hex bitstring support: x0ff
+              (setf (aref hexstr 0) #\X)
+              (loop :for pos :from 1 :by 2
+                 :for byte :across bytes
+                 :do  (let ((high (ldb (byte 4 4) byte))
+                            (low  (ldb (byte 4 0) byte)))
+                        (setf (aref hexstr pos)       (aref digits high))
+                        (setf (aref hexstr (+ pos 1)) (aref digits low))))
+              hexstr))))
+
 (defun int-to-ip (int)
   "Transform an IP as integer into its dotted notation, optimised code from
    stassats."

test/mysql/my.sql

@@ -155,6 +155,27 @@ create table `CamelCase` (
  `validSizes` varchar(12)
 );
 
+/*
+ * https://github.com/dimitri/pgloader/issues/943
+ */
+CREATE TABLE `countdata_template`
+(
+  `id` int(11) NOT NULL AUTO_INCREMENT,
+  `data` int(11) DEFAULT NULL,
+  `date_time` datetime DEFAULT NULL,
+  `gmt_offset` smallint(6) NOT NULL DEFAULT '0' COMMENT 'Offset GMT en minute',
+  `measurement_id` int(11) NOT NULL,
+  `flags` bit(16) NOT NULL DEFAULT b'0' COMMENT 'mot binaire : b1000=validé, b10000000=supprimé',
+  PRIMARY KEY (`id`),
+  UNIQUE KEY `ak_countdata_idx` (`measurement_id`,`date_time`,`gmt_offset`)
+)
+ENGINE=InnoDB DEFAULT CHARSET=utf8 COMMENT='données de comptage';
+
+INSERT INTO `countdata_template`(`date_time`, `measurement_id`, `flags`)
+     VALUES (now(), 1, b'1000'),
+            (now(), 2, b'10000000');
+
+
 CREATE TABLE `fcm_batches` (
   `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
   `raw_payload` mediumtext COLLATE utf8_unicode_ci,