Data type conversions

Data conversion in the template is two-fold:

the source data is converted to a corresponding Apache Beam data type
the Beam data is then converted to Neo4j data types

BigQuery types to Neo4j

BigQuery value conversion is driven by the BigQuery query result schema.

Property mapping types specified with target_property_type do not currently coerce BigQuery values before they are written to Neo4j. To change a BigQuery value type, cast or transform it in the source query.

BigQuery type Beam type Neo4j parameter value Notes

BigQuery type	Beam type	Neo4j parameter value	Notes
`BOOL`, `BOOLEAN`	`BOOLEAN`	`Boolean`
`INT64`, `INTEGER`	`INT64`	`Long`
`FLOAT64`, `FLOAT`	`DOUBLE`	`Double`
`NUMERIC`, `BIGNUMERIC`	`DECIMAL`	`Double`	Beam reads the value as `BigDecimal`; the template calls `doubleValue()`, so precision can be lost.
`STRING`	`STRING`	`String`
`BYTES`	`BYTES`	`byte[]`	BigQuery rows expose bytes as Base64 text and Beam decodes them before the Neo4j write.
`DATE`	Beam SQL `DATE` logical type	`java.time.LocalDate`
`TIME`	Beam SQL `TIME` logical type	`java.time.LocalTime`
`DATETIME`	Beam SQL `DATETIME` logical type	`java.time.LocalDateTime`
`TIMESTAMP`	`DATETIME` for standard microsecond timestamps	`java.time.OffsetDateTime` at UTC	The conversion uses epoch milliseconds, so sub-millisecond precision is not preserved.
`TIMESTAMP(12)`	Beam `Timestamp` logical type by default	`java.time.OffsetDateTime` at UTC	Beam’s default schema conversion maps picosecond timestamps to nanosecond precision. If Beam maps the value as `STRING`, the template writes it as a string.
`GEOGRAPHY`	`STRING`	`String`	The template does not convert WKT text to Neo4j `Point` values. Use custom Cypher® such as `point()` if you need points.
`JSON`	`STRING`	`String`	The JSON payload is written as text unless custom Cypher parses it.
`RECORD`, `STRUCT`	`ROW`	`Map<String,Object>`	Nested fields are recursively converted. This is useful only for custom Cypher query targets; node and relationship targets still only accept values Neo4j can store as properties.
`REPEATED` fields	`ARRAY` of the element Beam type	`List<Object>`	Element values are recursively converted. Repeated structs become lists of maps.
`RANGE`, `INTERVAL`	Not supported	Not written	Beam schema conversion rejects these types in the current template path.

BOOL, BOOLEAN

BOOLEAN

Boolean

INT64, INTEGER

INT64

Long

FLOAT64, FLOAT

DOUBLE

Double

NUMERIC, BIGNUMERIC

DECIMAL

Double

Beam reads the value as BigDecimal; the template calls doubleValue(), so precision can be lost.

STRING

String

BYTES

byte[]

BigQuery rows expose bytes as Base64 text and Beam decodes them before the Neo4j write.

DATE

Beam SQL DATE logical type

java.time.LocalDate

TIME

Beam SQL TIME logical type

java.time.LocalTime

DATETIME

Beam SQL DATETIME logical type

java.time.LocalDateTime

TIMESTAMP

DATETIME for standard microsecond timestamps

java.time.OffsetDateTime at UTC

The conversion uses epoch milliseconds, so sub-millisecond precision is not preserved.

TIMESTAMP(12)

Beam Timestamp logical type by default

java.time.OffsetDateTime at UTC

Beam’s default schema conversion maps picosecond timestamps to nanosecond precision. If Beam maps the value as STRING, the template writes it as a string.

GEOGRAPHY

STRING

String

The template does not convert WKT text to Neo4j Point values. Use custom Cypher® such as point() if you need points.

JSON

STRING

String

The JSON payload is written as text unless custom Cypher parses it.

RECORD, STRUCT

ROW

Map<String,Object>

Nested fields are recursively converted. This is useful only for custom Cypher query targets; node and relationship targets still only accept values Neo4j can store as properties.

REPEATED fields

ARRAY of the element Beam type

List<Object>

Element values are recursively converted. Repeated structs become lists of maps.

RANGE, INTERVAL

Not supported

Not written

Beam schema conversion rejects these types in the current template path.

When a Beam MAP value reaches the writer, only string keys are supported. Map values are recursively converted, and non-string keys fail to write.

Null values

BigQuery NULL values are passed through as Cypher null values.

For node and relationship targets, assigning Cypher null to a non-key property means Neo4j does not store that property. If the property already exists on a matched entity, it is removed. Key property values must not be null.

For custom Cypher query targets, null values are passed as query parameters; their effect depends on the Cypher query.