Streams back the contents of all rows in key order
The ``ColumnInfo`` message assigns an integer ``id`` for each column
and the ``id`` is present in each cell belonging to that column (this
is done in order to avoid sending the ``ColumnInfo`` with each ``Cell``).
Streams back the contents of all rows in key order
The ``ColumnInfo`` message assigns an integer ``id`` for each column
and the ``id`` is present in each cell belonging to that column (this
is done in order to avoid sending the ``ColumnInfo`` with each ``Cell``).
The column id starts from 0 and are incremented with each new column found.
The ids are only valid during one single dump.
The dumped data does not contain information on any table characteristics
such as (primary) key, partitioning or other storage options.
Imports a stream of rows
The table has to exist in order to load data into it.
As soon as the server detects an error with one of the written
rows, it will forcefully close the connection and send back an
early error message. The client should stop streaming and handle
the error.
Note that the erratic condition causes the connection to be closed
even if the ``Keep-Alive`` request header was enabled.
The error response is of type ``ExceptionMessage`` and contains a
detail message of type ``WriteRowsExceptionDetail`` that provides
the number of rows that were successfully written by the client.
The client can use this information to link the error message to a
row (i.e. the bad row is at position ``count + 1`` of the stream).
One possible error could be that the table has defined a (primary)
key and one of the loaded rows contains no value for one of the
columns of the key.
The table load will overwrite any data existing in the table with the
same key as the imported row.
The table load will not update the historgrams so a histogram rebuild
is required after the load.
Rebulds histograms - this is required after a table load.
Turrently the time interval passed in the request will be used to select the
partitions which will be rebuild - any partition overlapping with the interval will be rebuilt.
If the table is not partitioned by time, the histogram for the entire table will be rebuild.