dialect/sql/sqljson: fix sqlite haskey and add docs

doc/md/predicates.md

@@ -23,6 +23,7 @@ title: Predicates
   - =, !=, >, <, >=, <= on nested values (JSON path).
   - Contains on nested values (JSON path).
   - HasKey, Len&lt;P>
+  - `null` checks for nested values (JSON path).
 - **Optional** fields:
   - IsNil, NotNil
 
@@ -185,7 +186,8 @@ SELECT DISTINCT `users`.`id`, `users`.`age`, `users`.`name` FROM `users` JOIN `c
 SELECT DISTINCT `users`.`id`, `users`.`age`, `users`.`name` FROM `users` WHERE EXISTS (SELECT * FROM `cars` WHERE `cars`.`model` = 'Tesla' AND `users`.`id` = `cars`.`owner_id`)
 ```
 
-#### Get all pets where pet name contains a specific pattern 
+#### Get all pets where pet name contains a specific pattern
+
 The generated code provides the `HasPrefix`, `HasSuffix`, `Contains`, and `ContainsFold` predicates for pattern matching.
 However, in order to use the `LIKE` operator with a custom pattern, use the following example.
 
@@ -202,3 +204,65 @@ The above code will produce the following SQL query:
 ```sql
 SELECT DISTINCT `pets`.`id`, `pets`.`owner_id`, `pets`.`name`, `pets`.`age`, `pets`.`species` FROM `pets` WHERE `name` LIKE '_B%'
 ```
+
+## JSON predicates
+
+JSON predicates are not generated by default as part of the code generation. However, ent provides an official package
+named [`sqljson`](https://pkg.go.dev/entgo.io/ent/dialect/sql/sqljson) for applying predicates on JSON columns using the
+[custom predicates option](#custom-predicates).
+
+#### Compare a JSON value
+
+```go
+sqljson.ValueEQ(user.FieldData, data)
+
+sqljson.ValueEQ(user.FieldURL, "https", sqljson.Path("Scheme"))
+
+sqljson.ValueNEQ(user.FieldData, content, sqljson.DotPath("attributes[1].body.content"))
+
+sqljson.ValueGTE(user.FieldData, status.StatusBadRequest, sqljson.Path("response", "status"))
+```
+
+#### Check a presence of a JSON key
+
+```go
+sqljson.HasKey(user.FieldData, sqljson.Path("attributes", "[1]", "body"))
+
+sqljson.HasKey(user.FieldData, sqljson.DotPath("attributes[1].body"))
+```
+
+Note that, a key with the `null` literal as a value also matches this operation.
+
+#### Check JSON `null` literals
+
+```go
+sqljson.ValueIsNull(user.FieldData)
+
+sqljson.ValueIsNull(user.FieldData, sqljson.Path("attributes"))
+
+sqljson.ValueIsNull(user.FieldData, sqljson.DotPath("attributes[1].body"))
+```
+
+Note that, the `ValueIsNull` returns true if the value is JSON `null`,
+but not database `NULL`.
+
+#### Compare the length of a JSON array
+
+```go
+sqljson.LenEQ(user.FieldAttrs, 2)
+
+sql.Or(
+	sqljson.LenGT(user.FieldData, 10, sqljson.Path("attributes")),
+    sqljson.LenLT(user.FieldData, 20, sqljson.Path("attributes")),
+)
+```
+
+#### Check if a JSON value contains another value
+
+```go
+sqljson.ValueContains(user.FieldData, data)
+
+sqljson.ValueContains(user.FieldData, attrs, sqljson.Path("attributes"))
+
+sqljson.ValueContains(user.FieldData, code, sqljson.DotPath("attributes[0].status_code"))
+```