Repository navigation
The read_rows method in the bigtable client does not provide an interface to specify open/closed key ranges #3592
Description
The read_rows method of the Table class in the bigtable client is designed such that the results are inclusive of the start key but exclusive of the end key. That seems a reasonable choice, until one considers that keys may be designed to accumulate in reverse order (as recommended in the docs: https://cloud.google.com/bigtable/docs/schema-design-time-series). In that case, it is necessary to flip the natural order of the keys to process a query -i.e. if I query from time t0 to t1, I will create two corresponding keys k0 and k1, but k1 will be the start key and k0 the end key in the call to read_rows.
Currently, the specification of open and closed for the start and end key is hard-coded in the _create_row_request method, and that makes it difficult for an end user of the library to alter the default behavior. Consider adding an option to the read_rows method, or better yet, an attribute of a Table instance, such that queries can implement different behaviors with respect to the start and end keys.
Activity
robertlacok commentedon Jul 12, 2017
I also miss this feature. Would it be sensible to add two more optional arguments to the
read_rowsmethod,include_start_keyandinclude_end_key, which default toTrueandFalserespectively (current behavior), but they allow setting it manually?From what I see, that only requires adding a few more if statements here, and some docs, right?
Let me know, if it's good, I can pick it up.