跳到主要内容

How to define a secondary index

Overview

This guide provides instructions to define a secondary index for a multi-index table.

Reference

See the following code reference:

Before you begin

Make sure you have the following prerequisites in place:

  • An Antelope development environment, for details consult the Get Started Guide,
  • A multi-index table testtab along with its test_table data structure, its mandatory primary index, and the type alias definition test_table_t. Please see How To Define A Primary Index to set up these prerequisites.

Procedure

Complete the following steps to define a secondary index for the multi-index table testtab.

1. Extend The Multi-Index Data Structure

Add a second data member secondary, of type eosio::name, to the test_table data structure that defines the testtab data.

  struct [[eosio::table]] test_table {
// this data member stores a name for each row of the multi-index table
name test_primary;
+ name secondary;
// additional data stored in table row
uint64_t datum;
// mandatory definition for primary key getter
uint64_t primary_key( ) const { return test_primary.value; }
};

2. Add The Secondary Index Accessor Method

Add by_secondary() method, which is the index accessor method to the new data member added. The secondary index, that will be added in the next step, will index this new data structure data member.

  struct [[eosio::table]] test_table {
// this data member stores a name for each row of the multi-index table
name test_primary;
name secondary;
// additional data stored in table row
uint64_t datum;
// mandatory definition for primary key getter
uint64_t primary_key( ) const { return test_primary.value; }
+ uint64_t by_secondary( ) const { return secondary.value; }
};

3. Define The Secondary Index

In the test_table_t type definition, add the definition of the secondary index with the use of the eosio::indexed_by template. eosio::index_by needs two parameters: the name of the index, "secid"_n, and a function call operator which extracts the value from the secondary data member as an index key. The function call operator is defined with the use of eosio::const_mem_fun template which receives two parameters: the data structure test_table and the reference to the getter function member by_secondary.

-  typedef eosio::multi_index<"testtaba"_n, test_table> test_table_t;
+ typedef eosio::multi_index<"testtaba"_n, test_table, eosio::indexed_by<"secid"_n, eosio::const_mem_fun<test_table, uint64_t, &test_table::by_secondary>>> test_table_t;

For reference see below the full contract definition code with all the changes described above.

multi_index_example.hpp

#include <eosio/eosio.hpp>
using namespace eosio;

// multi-index example contract class
class [[eosio::contract]] multi_index_example : public contract {
public:
using contract::contract;

// contract class constructor
multi_index_example( name receiver, name code, datastream<const char*> ds ) :
// contract base class contructor
contract(receiver, code, ds),
// instantiate multi-index instance as data member (find it defined below)
testtab(receiver, receiver.value)
{ }

struct [[eosio::table]] test_table {
// this data member stores a name for each row of the multi-index table
name test_primary;
name secondary;
// additional data stored in table row
uint64_t datum;
// mandatory definition for primary key getter
uint64_t primary_key( ) const { return test_primary.value; }
uint64_t by_secondary( ) const { return secondary.value; }
};

// the multi-index type definition, for ease of use a type alias `test_table_t` is defined,
// based on the multi_index template type, parametarized with a random name, the
// test_table data structure, and the secondary index
typedef eosio::multi_index<"testtaba"_n, test_table, eosio::indexed_by<"secid"_n, eosio::const_mem_fun<test_table, uint64_t, &test_table::by_secondary>>> test_table_t;

// the multi-index table instance declared as a data member of type test_table_t
test_table_t testtab;

[[eosio::action]] void set( name user );
[[eosio::action]] void print( name user );

using set_action = action_wrapper<"set"_n, &multi_index_example::set>;
using print_action = action_wrapper<"print"_n, &multi_index_example::print>;
};

Now you have instantiated the testtab as a multi-index table which has a primary index defined for its test_primary data member and a secondary index for its secondary data member.

Full example location

A full example project demonstrating the instantiation and usage of multi-index table can be found here.

Do not add a secondary index to an existing table

Adding a secondary index to an existing multi-index table it will have unpredictable outcome. Consult the Data design and migration documentation for more details.

Summary

In conclusion, the above instructions show how to define a secondary index for a multi-index table.

Next Steps