Using Casts in Secondary Indexes

Defining an index that specifies index keys with CAST functions provides fast access for queries that contain CAST functions. The index converts the indexed field to the type specified by the CAST function and stores the result.

Create indexes using CAST functions if you want to CAST fields to specific data types in your queries. To define an index with the CAST function applied to a field, specify a CAST when defining the index key. The following example creates an index that casts the age field to an INT type and the height field to a FLOAT type. 
maprcli table index add -path /castTable -index castIdx \
   -indexedfields '$CAST(age@INT)','$CAST(height@FLOAT)'

When issuing Drill queries through Business Intelligence (BI) tools, you can include CAST functions in your queries to create Drill views. Including CAST functions provides the metadata needed to optimally process the queries. For more information about using the CAST function with Drill, see Data Type Conversion.

Casting from NULL in Drill

You can cast from null to any data type supported by the indexes with the CAST function. However, null can be a valid JSON value for the string data type, for example:
{ "name": null }
Null can also represent the absence of an actual value, for example:
{ "_id":1,"name":"Annie" }
{ "_id":2 }  (name does not exist)

When you cast on columns with missing values, Drill does not return null for the missing values. Drill only returns null in cases where an actual value of null exists.

For example, if you have the following data stored in a JSON table named t1:
a1 b1
1 'abc'
2 null
3 'null'
And you issue the following query against the table:
SELECT a1, b1(cast b1 as varchar(20))) from t1;
Drill returns the following data:
a1 b1
1 abc
2
3 null
Drill does not return null where null represents a missing value. Drill only returns null in the instance where null is stored as a string value.

Guidelines for Using Casts in Indexes

The following rules apply to CAST functions used in secondary indexes:

  • You can include the CAST function only on indexed fields.
  • You do not have to cast between comparable data types.
  • Indexes support casting to the following data types:
    • Boolean
    • String
    • Int
    • Long
    • Float
    • Double
    • Date
    • Time
    • Timestamp
    NOTE MapR Database does not support casting from any data type to byte, short, decimal, binary, or interval.
    NOTE Queries that use the CAST function on fields with timestamp and binary data types are not supported.
  • When casting to a string type, you can optionally specify a length. If you do not specify a length, it defaults to the maximum length of 255.
  • When casting a float or double type to a string type, you cannot control the precision of the digits in the resulting string value. Float and double are approximate representations of decimal values.
  • When casting from a binary type, MapR Database assumes that the binary value is a UTF-8 formatted string representation of the resulting data type.
  • If MapR Database cannot cast a value, MapR Database indexes the row with an encoding error that specifies a CAST issue.
  • You cannot cast all data types to the supported data types. See the Casting Matrix below for supported and unsupported combinations.

Casting Matrix

The following matrix displays supported and unsupported casting, from the data type shown in the column to the data type shown in the row. Y indicates a supported casting; N indicates an unsupported casting. Hyphen (-) indicates that casting is unnecessary, because the data types are comparable.
int long float double string boolean date time timestamp
byte Y Y Y Y Y Y N N N
short Y Y Y Y Y Y N N N
int - Y Y Y Y Y Y1 Y2 Y
long Y - Y Y Y Y Y Y Y
float Y Y - Y Y Y N N N
double Y Y Y - Y Y N N N
string Y Y Y Y - Y3 Y Y Y
boolean Y Y N N Y - N N N
date N N N N Y N - Y Y
time N N N N Y N N - N
timestamp N N N N Y4 N Y Y -
binary Y Y Y Y Y N N N N
array N N N N N N N N N
nested document N N N N N N N N N

1 When casting int/long to a date type, the date value is constructed based on the int/long value being the number of milliseconds since epoch.

2 When casting int/long to a time type, the time value is constructed based on the int/long value being the time of day in milliseconds.

3 MapR Database casts the strings true, yes, on, y, t, and 1 to boolean true. MapR Database casts the strings false, no, off, n, f, and 0 to boolean false.

4 The string represents the time in UTC timezone.

Example Using Cast Function in an Index

This example shows you how to create an index with the CAST function.

The following statement queries a table named lineitem and casts the L_LINENUMBER and L_ORDERKEY fields to the int data type:
SELECT L_LINESTATUS, L_QUANTITY FROM lineitem WHERE CAST(L_LINENUMBER as int) = 1 AND CAST(L_ORDERKEY as int) = 550;
You can create an index on the L_LINENUMBER and L_ORDERKEY fields and indicate the use of the CAST function and data type for each field, as follows:
maprcli table index add -path /drill/testdata/qa/sf1/maprdb/json/lineitem -index l_cast_comp_1 \
   -indexedfields '$CAST(L_LINENUMBER@INT)','$CAST(L_ORDERKEY@INT)' \
   -includedfields L_LINESTATUS,L_QUANTITY
The index stores the values of the L_LINENUMBER and L_ORDERKEY fields as the int data type. MapR Database can use the index for any subsequent queries that use the CAST function to retrieve these fields as the int type, instead of accessing data in the primary table and converting the values to int.
NOTE If you created an index on the L_LINENUMBER and L_ORDERKEY fields without the CAST function, the query used in this example would not benefit from the index.