AnyDAC allows to insert a new record with an auto-incrementing column and get back a new value of this column. That works as for immediate updates, as for cached updates. Depending on a DBMS, the auto-incrementing fields may be implemented either using a special IDENTITY (or similar) column data type, either using a generator (or a sequence) and a table trigger:

AnyDAC automatically recognizes a column of an auto-incrementing data type, and defines it as dtIntXxx, [caAutoInc, caReadOnly, caAllowNull]. This leads to TField setup:

• TField.DataType = ftAutoInc (TADAutoIncField) when dtInt32 or dtUInt32; otherwise one of the numeric ftXxxx data types;
• TField.Required = False;
• TField.ReadOnly = True;
• TField.ProviderFlags = [pfInWhere], or [pfInWhere, pfInKey] when the column is part of a primary key.

AnyDAC automatically recognizes limited set of the Firebird auto-incrementing columns, and defines them as dtIntXxx, [caAutoInc, caAllowNull] when:

• extended metadata is enabled;
• a table has a BEFORE INSERT trigger;
• the trigger depends on a single column and a single generator. This column is recognizing as auto-incrementing.

For Oracle, other Firebird / Interbase cases, and other columns the auto-incrementing mode may be specified by programmer, using one of the options:

• set UpdateOptions.AutoIncFields to the list of the auto-incrementing column names. When a column is of dtInt32 / dtUInt32 data type, then TADAutoIncField field will be created.
• create an TADAutoIncField field at design or run time. The column must be of ftAutoInc / ftInteger / ftLongWord field type. For other data types consider to use data type mapping to TargetDataType = dtInt32.
• set TField.AutoGenerateValue to arAutoInc for an auto-incrementing field. This method does not create TADAutoIncField, works with any field type, and may require additional field properties setup, eg setting ProviderFlags, Required and ReadOnly.

All these methods at end leads to TField.AutoGenerateValue = arAutoInc. 

By default AnyDAC uses the client side auto-incrementing for the auto-incrementing columns. After calling dataset Insert / Append method, the auto-incrementing column will get -1 value. With each next method call, the value will be incremented with the -1 (negative) step. That was made to distinguish the client assigned values and the DBMS assigned values. 

When UpdateOptions.RefreshMode <> rmManual, then after posting a new record the auto-incrementing column will get an actual positive value. 

The TADAutoIncField properties allows to adjust the client side auto-incrementing:

• ClientAutoIncrement to enable or disable;
• AutoIncrementSeed to specify an initial value;
• AutoIncrementStep to specify a step.

When a DBMS supports sequences or generators (here they are synonyms), then AnyDAC may fill an auto-incrementing column from a sequence on a client. This is alternative to client-side auto-incrementing. 

The auto-incrementing column must have pfInUpdate in TField.ProviderFlags and one of the UpdateOptions.FetchGeneratorsPoint values:

• gpImmediate - a next value is fetched from a generator in TDataSet.Insert / Append method. And the value is already accessible in TDataSet.OnNewRecord event handler. When the TDataSet.Cancel method is called, then the value is lost.
• gpDeferred - a next value is fetched in TDataSet.Post method.

Set the UpdateOptions.GeneratorName or TADAutoIncField.GeneratorName to a generator name, which will be used to get next sequence value. 

The following auto-incrementing column setup will work with any DBMS, including supporting native auto-incremental columns, and using sequences and triggers:

• UpdateOptions.FetchGeneratorsPoint = gpNone. The client sequence filling is disabled.
• not (pfInUpdate in TField.ProviderFlags). The column is excluded from INSERT / UPDATE clause.
• fiMeta in FetchOptions.Items. The column, when it is possible is recognized as auto-incremental.

For Oracle, Firebird and Interbase is required to manually specify auto-incremental columns. For databases supporting native auto-incremental data types, no additional setup is required. But specifying will work in both cases. 

When master and detail datasets are linked by an auto-incrementing master column and cached updates is enabled, then consider to use Centralized Cached Updates mode with propagation of changes from master to detail datasets.

Unique Identifying Fields, Update Command Generation, Caching Updates, FetchGeneratorsPoint, GeneratorName, AutoIncFields, TADAutoIncField