- Singer-Python: bumped to latest
- Minor housekeeping:
- Updated container versions to latest
- Updated README to reflect new versions of PostgreSQL Server
- BUG FIX: A bug was identified for de-nesting.
- ISSUE LINK
- FAILING TESTS LINK
- FIX LINK
- Subtables with subtables did not serialize column names correctly
- The column names ended up having the table names (paths) prepended on them
- Due to the denested table schema and denested records being different no information showed up in remote.
- This bug was ultimately tracked down to the core denesting logic.
- This will fix failing uploads which had nullable columns in subtables but
no data was seen populating those columns.
- The broken schema columns will still remain
- Failing schemas which had non-null columns in subtables will still be broken
- To fix will require dropping the associated tables, potentially resetting the entire
db
/schema
- To fix will require dropping the associated tables, potentially resetting the entire
- BUG FIX: A bug was identified for path to column serialization.
- LINK
- A nullable properties which had multiple JSONSchema types
- ie, something like
[null, string, integer ...]
- Failed to find an appropriate column in remote to persist
None
values to.
- ie, something like
- Found by usage of the Hubspot Tap
- FEATURES:
- Added the
persist_empty_tables
config option which allows the Target to create empty tables in Remote.
- Added the
- BUG FIX: A bug was identified in 0.1.3 with stream
key_properties
and canonicalization.- LINK
- Discovered and fixed by @mirelagrigoras
- If the
key_properties
for a stream changed due to canonicalization, the stream would fail to persist due to:- the
persist_csv_rows
key_properties
values would remain un-canonicalized (sp?) and therefore cause issues once serialized into a SQL statement - the pre-checks for tables would break because no values could be pulled from the schema with un-canonicalized fields pulled out of the
key_properties
- the
- NOTE: the
key_properties
metadata is saved with raw field names.
- SCHEMA_VERSION: 1
- LINK
- Initialized a new field in remote table schemas
schema_version
- A migration in
PostgresTarget
handles updating this
- BUG FIX: A bug was identified in 0.1.2 with column type splitting.
- LINK
- A schema with a field of type
string
is persisted to remote- Later, the same field is of type
date-time
- The values for this field will not be placed under a new column, but rather under the original
string
column
- The values for this field will not be placed under a new column, but rather under the original
- Later, the same field is of type
- A schema with a field of type
date-time
is persisted to remote- Later, the same field is of type
string
- The original
date-time
column will be madenullable
- The values for this field will fail to persist
- The original
- Later, the same field is of type
- FEATURES:
- Added the
logging_level
config option which uses standard Python Logger Levels to configure more details about what Target-Postgres is doing- Query level logging and timing
- Table schema changes logging and timing
- Added the