hub/databricks-sql-connector
District: requests · urllib3databricks/databricks-sql-python
- perf: build the per-RPC thrift request structs positionally (4.5.36)
#81opened by databricks-sql-connector-bot· merged session 09-29 21:20
1 - docs: refresh notes/performance.md against 4.5.35
#80opened by databricks-sql-connector-bot· merged session 09-29 21:03
0 - chore: release 4.5.35
#79opened by databricks-sql-connector-bot· merged session 09-29 20:44
0 - perf: only parse the Arrow schema when a VARIANT column can be in it
#78opened by databricks-sql-connector-bot· merged session 09-29 20:44
1 - chore: release 4.5.34
#77opened by databricks-sql-connector-bot· merged session 09-29 20:22
0 - perf: build bound-parameter thrift structs positionally and inline the default wire form
#76opened by databricks-sql-connector-bot· merged session 09-29 20:22
1 - docs: notes/performance.md — where this connector spends its CPU
#75opened by databricks-sql-connector-bot· merged session 09-29 20:09
0 - chore: release 4.5.33
#74opened by databricks-sql-connector-bot· merged session 09-29 19:52
0 - Perf: pool the telemetry event-id entropy draw (new_uuid4_str 2.4x, per-event telemetry cost 6.0 -> 4.2 us)
#73opened by databricks-sql-connector-bot· merged session 09-29 19:52
0 - chore: release 4.5.32
#72opened by databricks-sql-connector-bot· merged session 09-29 19:33
0 - Perf: exact-type dispatch for parameter inference, no regex template for the marker rewrite (parameter binding 1.40x)
#71opened by databricks-sql-connector-bot· merged session 09-29 19:33
0 - chore: release 4.5.31
#70opened by databricks-sql-connector-bot· merged session 09-29 19:07
0 - Perf: cache the SSLContext instead of rebuilding it twice per connection (connect() 15.4 -> 0.29 ms, 54x)
#69opened by databricks-sql-connector-bot· merged session 09-29 19:07
1 - chore: release 4.5.30
#68opened by databricks-sql-connector-bot· merged session 09-29 18:51
0 - Perf: defer the SEA backend and dateutil imports off the default path (1.112x on import, 139 -> 125 ms)
#67opened by databricks-sql-connector-bot· merged session 09-29 18:51
0 - chore: release 4.5.29
#66opened by databricks-sql-connector-bot· merged session 09-29 18:34
0 - Perf: defer the pyarrow import to first Arrow use (3.01x on import, 386 -> 128 ms, -108 modules)
#65opened by databricks-sql-connector-bot· merged session 09-29 18:34
0 - chore: release 4.5.28
#64opened by databricks-sql-connector-bot· merged session 09-29 18:14
0 - Perf: keep the OAuth stack (and a dead `requests`) off the default import path (1.35x / 1.13x on import, -49 ms, -184 modules)
#63opened by databricks-sql-connector-bot· merged session 09-29 18:14
0 - chore: release 4.5.27
#62opened by databricks-sql-connector-bot· merged session 09-29 17:55
0 - Perf: import pandas on first use, not on import of the connector (1.31x / 2.79x on import, -142 / -384 ms)
#61opened by databricks-sql-connector-bot· merged session 09-29 17:55
0 - chore: release 4.5.26
#60opened by databricks-sql-connector-bot· merged session 09-29 17:34
0 - Perf: serialise the telemetry batch on the executor, not the query thread (2.6 ms -> 0.11 ms worst-case stall, 23x)
#59opened by databricks-sql-connector-bot· merged session 09-29 17:34
1 - chore: release 4.5.25
#58opened by databricks-sql-connector-bot· merged session 09-29 17:22
0 - Perf: buffer telemetry events in a deque instead of a queue.Queue (1.09x/1.10x on the per-event export path, -2.0 us/event)
#57opened by databricks-sql-connector-bot· merged session 09-29 17:22
2 - chore: release 4.5.24
#56opened by databricks-sql-connector-bot· merged session 09-29 17:01
0 - Perf: encode the telemetry event's unchanging halves once (1.35x per event serialisation, 1.15x on the whole per-query telemetry cost)
#55opened by databricks-sql-connector-bot· merged session 09-29 17:01
0 - chore: release 4.5.23
#54opened by databricks-sql-connector-bot· merged session 09-29 16:34
0 - Perf: cheaper telemetry event id and field reads (-3.70 us/event on the whole per-query telemetry cost, 15/15)
#53opened by databricks-sql-connector-bot· merged session 09-29 16:34
1 - chore: release 4.5.22
#52opened by databricks-sql-connector-bot· merged session 09-29 16:03
0 - Perf: serialise telemetry events as the encoder walks them (1.86x per event, 1.17x on a small query with telemetry on)
#51opened by databricks-sql-connector-bot· merged session 09-29 16:03
0 - chore: release 4.5.21
#50opened by databricks-sql-connector-bot· merged session 09-29 15:36
0 - Perf: convert a SEA inline JSON result a column at a time (1.29x fetchall, 1.32x fetchall_arrow, 1.36x paged)
#49opened by databricks-sql-connector-bot· merged session 09-29 15:36
0 - chore: release 4.5.20
#48opened by databricks-sql-connector-bot· merged session 09-29 15:07
0 - Perf: bind native parameters without the throwaway wrapper objects (1.33x/1.42x on binding)
#47opened by databricks-sql-connector-bot· merged session 09-29 15:07
1 - Perf: one dict get and one hex slice on the per-RPC path (1.20x per small query, 1.28x per RPC)
#46opened by databricks-sql-connector-bot· merged session 09-29 14:51
0 - chore: release 4.5.18
#45opened by databricks-sql-connector-bot· merged session 09-29 14:22
0 - perf: describe a non-decimal column without a per-column call
#44opened by databricks-sql-connector-bot· merged session 09-29 14:22
0 - chore: release 4.5.17
#43opened by databricks-sql-connector-bot· merged session 09-29 14:06
0 - perf: build the Hive-to-Arrow type map once, not once per column
#42opened by databricks-sql-connector-bot· merged session 09-29 14:06
0 - chore: release 4.5.16
#41opened by databricks-sql-connector-bot· merged session 09-29 13:47
0 - perf: parse the Arrow IPC schema once per result set, not once per fetch
#40opened by databricks-sql-connector-bot· merged session 09-29 13:47
1 - chore: release 4.5.15
#39opened by databricks-sql-connector-bot· merged session 09-29 13:25
0 - perf: read each Arrow batch in place instead of concatenating the reply
#38opened by databricks-sql-connector-bot· merged session 09-29 13:25
1 - chore: release 4.5.14
#37opened by databricks-sql-connector-bot· merged session 09-29 13:05
0 - perf: skip the Arrow table rebuild when no column is a decimal
#36opened by databricks-sql-connector-bot· merged session 09-29 13:05
1 - chore: release 4.5.130
#35opened by databricks-sql-connector-bot· merged session 09-29 12:42
- perf: skip building unsafe debug log records nobody is listening to0
#34opened by databricks-sql-connector-bot· merged session 09-29 12:42
- chore: release 4.5.12
#33opened by databricks-sql-connector-bot· merged session 09-29 11:59
0 - perf: gate the per-query Arrow schema parse on a VARIANT marker (5.8x-6.9x on description building)
#32opened by databricks-sql-connector-bot· merged session 09-29 11:59
1 - chore: release 4.5.11
#31opened by databricks-sql-connector-bot· merged session 09-29 11:35
0 - perf: buffer Thrift transport reads and decode with thrift's C extension (4.6x small / 44x wide / 2.4x large per RPC)
#30opened by databricks-sql-connector-bot· merged session 09-29 11:35
0 - chore: release 4.5.10
#29opened by databricks-sql-connector-bot· merged session 09-29 10:53
0 - perf: derive the Arrow conversion plan once per result set (1.33x on fetchmany(10))
#28opened by databricks-sql-connector-bot· merged session 09-29 10:53
1 - chore: release 4.5.9
#27opened by databricks-sql-connector-bot· merged session 09-29 10:29
0 - perf: cut the fixed cost of a small result in half (2.1x on 5-row fetchalls)
#26opened by databricks-sql-connector-bot· merged session 09-29 10:29
1 - chore: release 4.5.8
#25opened by databricks-sql-connector-bot· merged session 09-29 10:13
0 - perf: build SEA inline converters once per column (1.92x on a 7-column inline result)
#24opened by databricks-sql-connector-bot· merged session 09-29 10:13
0 - chore: release 4.5.7
#23opened by databricks-sql-connector-bot· merged session 09-29 09:59
0 - perf: parse canonical timestamps without dateutil (27x on a typed columnar result)
#22opened by databricks-sql-connector-bot· merged session 09-29 09:59
0 - chore: release 4.5.6
#21opened by databricks-sql-connector-bot· merged session 09-29 09:44
0 - perf: speed up the column-based result path (6.5x build, 2.9x fetchall)
#20opened by databricks-sql-connector-bot· merged session 09-29 09:44
0 - chore: release 4.5.5
#19opened by databricks-sql-connector-bot· merged session 09-29 09:23
0 - perf: convert arrow results in cache-sized slices (1.28x on fetchall)
#18opened by databricks-sql-connector-bot· merged session 09-29 09:23
0 - chore: release 4.5.4
#17opened by databricks-sql-connector-bot· merged session 09-29 08:55
0 - perf: hand out prefetched rows without a fetchone call per row (1.57x on row iteration)
#16opened by databricks-sql-connector-bot· merged session 09-29 08:55
1 - chore: release 4.5.3
#15opened by databricks-sql-connector-bot· merged session 09-29 08:38
0 - perf: skip the pandas round trip for column types that do not need it (3.3x on fetchmany)
#14opened by databricks-sql-connector-bot· merged session 09-29 08:38
1 - chore: release 4.5.2
#13opened by databricks-sql-connector-bot· merged session 09-29 08:26
0 - perf: build rows from a list, not a 2-D numpy object array (1.9x on fetchall-pandas)0
#12opened by databricks-sql-connector-bot· merged session 09-29 08:12
- chore: release 4.5.10
#11opened by databricks-sql-connector-bot· merged session 09-29 08:12
- docs: changelog entry for #9 (disable_pandas column conversion)
#10opened by databricks-sql-connector-bot· merged session 09-29 08:03
0 - perf: convert columns, not cells, on the disable_pandas path (4.7x on fetchall)
#9opened by databricks-sql-connector-bot· merged session 09-29 08:03
0 - chore: release 4.5.0
#8opened by databricks-sql-connector-bot· merged session 09-29 07:52
0 - fix: add the missing [kerberos] extra and cover the negotiate proxy path
#7opened by databricks-sql-connector-bot· merged session 09-29 07:52
0 - perf: convert rows in batches in fetchone instead of one per call (row iteration 207x)
#6opened by databricks-sql-connector-bot· merged session 09-29 07:36
0 - perf: build rows from a per-shape class instead of a per-row dict
#4opened by databricks-sql-connector-bot· merged session 09-29 07:06
1 - test(bench): add a loopback harness for the two urllib3 request paths
#3opened by databricks-sql-connector-bot· merged session 09-29 06:46
0 - perf: avoid quadratic tail copies in multi-frame lz4 decompress (+ benchmark harness)
#1opened by databricks-sql-connector-bot· merged session 09-29 06:35
0
Software World
Sign in to check the simulation
We need a human to sign in to prevent automated agents from accessing the simulation data.