v2.4.0

Compare Source

Features

• Add "dayofyear" property for dt accessors (#​1692) (9d4a59d)
• Add .dt.days, .dt.seconds, dt.microseconds, and dt.total_seconds() for timedelta series. (#​1713) (2b3a45f)
• Add DatetimeIndex class (#​1719) (c3c830c)
• Add isocalendar() for dt accessor" (#​1717) (0479763)
• Add bigframes.bigquery.json_value (#​1697) (46a9c53)
• Add blob.exif function support (#​1703) (3f79528)
• Add inplace arg support to sort methods (#​1710) (d1ccb52)
• Improve error message in Series.apply for direct udfs (#​1673) (1a658b2)
• Publish bigframes blob(Multimodal) to preview (#​1693) (e4c85ba)
• Support () operator between timedeltas (#​1702) (edaac89)
• Support forecast_limit_lower_bound and forecast_limit_upper_bound in ARIMA_PLUS (and ARIMA_PLUS_XREG) models (#​1305) (b16740e)
• Support to_strip parameter for str.strip, str.lstrip and str.rstrip (#​1705) (a84ee75)

Bug Fixes

• Fix dayofyear doc test (#​1701) (9b777a0)
• Fix issues with chunked arrow data (#​1700) (e3289b7)
• Rename columns with protected names such as _TABLE_SUFFIX in to_gbq() (#​1691) (8ec6079)

Performance Improvements

• Defer query in read_gbq with wildcard tables (#​1661) (5c125c9)
• Rechunk result pages client side (#​1680) (67d8760)

Dependencies

• Move bigtable and pubsub to extras (#​1696) (597d817)

Documentation

• Add snippets for Matrix Factorization tutorials (#​1630) (24b37ae)
• Deprecate bpd.options.bigquery.allow_large_results in favor of bpd.options.compute.allow_large_results (#​1597) (18780b4)
• Include import statement in the bigframes code snippet (#​1699) (08d70b6)
• Include the clean-up step in the udf code snippet (#​1698) (48992e2)
• Move multimodal notebook out of experimental folder (#​1712) (68b6532)
• Update blob_display option in snippets (#​1714) (8b30143)

v2.3.0

Compare Source

Features

• Add dry_run parameter to read_gbq(), read_gbq_table() and read_gbq_query() (#​1674) (4c5dee5)

Bug Fixes

• Guarantee guid thread safety across threads (#​1684) (cb0267d)
• Support large lists of lists in bpd.Series() constructor (#​1662) (0f4024c)
• Use value equality to check types for unix epoch functions and timestamp diff (#​1690) (81e8fb8)

Performance Improvements

• to_datetime() now avoids caching inputs unless data is inspected to infer format (#​1667) (dd08857)

Documentation

• Add a visualization notebook to BigFrame samples (#​1675) (ee062bf)
• Fix spacing of k-means code snippet (#​1687) (99f45dd)
• Update snippet for Create a k-means model tutorial (#​1664) (761c364)

v2.2.0

Compare Source

Features

• Add gemini-2.0-flash-001 and gemini-2.0-flash-lite-001 to fine tune score endponts and multimodal endpoints (#​1650) (4fb54df)
• Add GeminiTextGenerator.predict structured output (#​1653) (6199023)
• DataFrames.getitem support for slice input (#​1668) (563f0cb)
• Print right origin of PreviewWarning for the bpd.udf (#​1629) (48d10d1)
• Session.bytes_processed_sum will be updated when allow_large_re… (#​1669) (ae312db)
• Short circuit query for local scan (#​1618) (e84f232)
• Support names parameter in read_csv for bigquery engine (#​1659) (3388191)
• Support passing list of values to bigframes.core.sql.simple_literal (#​1641) (102d363)
• Support write api as loading option (#​1617) (c46ad06)

Bug Fixes

• DataFrame accessors is not pupulated (#​1639) (28afa2c)
• Prefer remote schema instead of throwing on materialize conflicts (#​1644) (53fc25b)
• Remove itertools.pairwise usage (#​1638) (9662745)
• Resolve issue where pre-release versions of google-auth are installed (#​1491) (ebb7a5e)
• Resolve some of the typo errors (#​1655) (cd7fbde)

Performance Improvements

• Fold row count ops when known (#​1656) (c958dbe)
• Use flyweight for node fields (#​1654) (8482bfc)

Dependencies

• Support shapely 1.8.5+ again (#​1651) (ae83e61)

Documentation

• Add JSON data types notebook (#​1647) (9128c4a)
• Add sample code snippets for udf (#​1649) (53caa8d)
• Fix bq_dataframes_template notebook to work if partial ordering mode is enabled (#​1665) (f442e7a)
• Note that udf is in preview and must be python 3.11 compatible (#​1629) (48d10d1)

v2.1.0

Compare Source

Features

• Add bigframes.bigquery.st_distance function (#​1637) (bf1ae70)
• Enable local json string validations (#​1614) (233347a)
• Enhance read_csv index_col parameter support (#​1631) (f4e5b26)

Bug Fixes

• Add retry for test_clean_up_via_context_manager (#​1627) (58e7cb0)
• Improve robustness of managed udf code extraction (#​1634) (8cc56d5)

Documentation

• Add code samples in the udf API docstring (#​1632) (f68b80c)

v2.0.0

Compare Source

⚠ BREAKING CHANGES

• make dataset and name params mandatory in udf (#​1619)
• Locational endpoints support is not available in BigFrames 2.0.
• change default LLM model to gemini-2.0-flash-001, drop PaLM2TextGenerator and PaLM2TextEmbeddingGenerator (#​1558)
• change default ingress setting for remote_function to internal-only (#​1544)
• make remote_function params keyword only (#​1537)
• make remote_function default service account explicit (#​1537)
• set allow_large_results=False by default (#​1541)

Features

• Add on parameter in dataframe.rolling() and dataframe.groupby.rolling() (#​1556) (45c9d9f)
• Add component to manage temporary tables (#​1559) (0a4e245)
• Add Series.to_pandas_batches() method (#​1592) (09ce979)
• Add support for creating a Matrix Factorization model (#​1330) (b5297f9)
• Allow input_types, output_type, and dataset to be used positionally in remote_function (#​1560) (bcac8c6)
• Allow pandas.cut 'labels' parameter to accept a list of string (#​1549) (af842b1)
• Change default ingress setting for remote_function to internal-only (#​1544) (c848a80)
• Detect duplicate column/index names in read_gbq before send query. (#​1615) (40d6960)
• Drop support for locational endpoints (#​1542) (4bf2e43)
• Enable time range rolling for DataFrame, DataFrameGroupBy and SeriesGroupBy (#​1605) (b4b7073)
• Improve local data validation (#​1598) (815e471)
• Make remote_function default service account explicit (#​1537) (9eb9089)
• Set allow_large_results=False by default (#​1541) (e9fb712)
• Support bigquery connection in managed function (#​1554) (f6f697a)
• Support bq connection path format (#​1550) (e7eb918)
• Support gemini-2.0-X models (#​1558) (3104fab)
• Support inlining small list, struct, json data (#​1589) (2ce891f)
• Support time range rolling on Series. (#​1590) (6e98a2c)
• Use session temp tables for all ephemeral storage (#​1569) (9711b83)
• Use validated local storage for data uploads (#​1612) (aee4159)
• Warn the deprecated max_download_size, random_state and sampling_method parameters in (DataFrame|Series).to_pandas() (#​1573) (b9623da)

Bug Fixes

• to_pandas_batches() respects page_size and max_results again (#​1572) (27c5905)
• Ensure page_size works correctly in to_pandas_batches when max_results is not set (#​1588) (570cff3)
• Include role and service account in IAM exception (#​1564) (8c50755)
• Make dataset and name params mandatory in udf (#​1619) (637e860)
• Pandas.cut returns labels index for numeric breaks when labels=False (#​1548) (b2375de)
• Prevent KeyError in bpd.concat with empty DF and struct/array types DF (#​1568) (b4da1cf)
• Read_csv supports for tilde local paths and includes index for bigquery_stream write engine (#​1580) (352e8e4)
• Use dictionaries to avoid problematic google.iam namespace (#​1611) (b03e44f)

Performance Improvements

• Directly read gbq table for simple plans (#​1607) (6ad38e8)

Dependencies

• Remove jellyfish dependency (#​1604) (1ac0e1e)
• Remove parsy dependency (#​1610) (293f676)
• Remove test dependency on pytest-mock package (#​1622) (1ba72ea)
• Support a shapely versions 1.8.5+ (#​1621) (e39ee3b)

Documentation

• Add details for bigquery_connection in [@bpd](https://redirect.github.com/bpd).udf docstring (#​1609) (ef63772)
• Add explain forecast snippet to multiple time series tutorial (#​1586) (40c55a0)
• Add message to remove default model for version 3.0 (#​1563) (910be2b)
• Add samples for ArimaPlus time_series_id_col feature (#​1577) (1e4cd9c)
• Add warning for bigframes 2.0 (#​1557) (3f0eaa1)
• Deprecate default model in TextEmbedddingGenerator, GeminiTextGenerator, and other bigframes.ml.llm classes (#​1570) (89ab33e)
• Include all licenses for vendored packages in the root LICENSE file (#​1626) (8116ed0)
• Remove gemini-1.5 deprecation warning for GeminiTextGenerator (#​1562) (0cc6784)
• Use restructured text to allow publishing to PyPI (#​1565) (d1e9ec2)

Miscellaneous Chores

• Make remote_function params keyword only (#​1537) (9eb9089)

v1.42.0

Compare Source

Features

• Add closed parameter in rolling() (#​1539) (8bcc89b)
• Add GeoSeries.difference() and bigframes.bigquery.st_difference() (#​1471) (e9fe815)
• Add GeoSeries.intersection() and bigframes.bigquery.st_intersection() (#​1529) (8542bd4)
• Add df.take and series.take (#​1509) (7d00be6)
• Add Linear_Regression.global_explain() (#​1446) (7e5b6a8)
• Allow iloc to support lists of negative indices (#​1497) (a9cf215)
• Support dry_run in to_pandas() (#​1436) (75fc7e0)
• Support window partition by geo column (#​1512) (bdcb1e7)
• Upgrade BQ managed udf to preview (#​1536) (4a7fe4d)

Bug Fixes

• Add deprecation warning to TextEmbeddingGenerator model, espeically gemini-1.0-X and gemini-1.5-X (#​1534) (c93e720)
• Change the default value for pdf extract/chunk (#​1517) (a70a607)
• Local data always has sequential index (#​1514) (014bd33)
• Read_pandas inline returns None when exceeds limit (#​1525) (578081e)
• Temporary fix for StreamingDataFrame not working backend bug (#​1533) (6ab4ffd)
• Tolerate BQ connection service account propagation delay (#​1505) (6681f1f)

Performance Improvements

• Update shape to use quer_and_wait (#​1519) (34ab9b8)

Documentation

• Update GeoSeries.difference() and bigframes.bigquery.st_difference() docs (#​1526) (d553fa2)

v1.41.0

Compare Source

Features

• Add support for the 'right' parameter in 'pandas.cut' (#​1496) (8aff128)
• Support BQ managed functions through read_gbq_function (#​1476) (802183d)
• Warn when the BigFrames version is more than a year old (#​1455) (00e0750)

Bug Fixes

• Fix pandas.cut errors with empty bins (#​1499) (434fb5d)
• Fix read_gbq with ORDER BY query and index_col set (#​963) (de46d2f)

Performance Improvements

• Eliminate count queries in llm retry (#​1489) (1c934c2)

Documentation

• Add a sample notebook for vector search (#​1500) (f3bf139)

v1.40.0

Compare Source

⚠ BREAKING CHANGES

• reading JSON data as a custom arrow extension type (#​1458)

Features

• Reading JSON data as a custom arrow extension type (#​1458) (e720f41)
• Support list output for managed function (#​1457) (461e9e0)

Bug Fixes

• Fix list-like indexers in partial ordering mode (#​1456) (fe72ada)
• Fix the merge issue between 1424 and 1373 (#​1461) (7b6e361)
• Use == instead of is for timedelta type equality checks (#​1480) (0db248b)

Performance Improvements

• Compilation no longer bounded by recursion (#​1464) (27ab028)

v1.39.0

Compare Source

Features

• (Preview) Support diff() for date series (#​1423) (521e987)
• (Preview) Support aggregations over timedeltas (#​1418) (1251ded)
• (Preview) Support arithmetics between dates and timedeltas (#​1413) (962b152)
• (Preview) Support automatic load of timedelta from BQ tables. (#​1429) (b2917bb)
• Add allow_large_results option to many I/O methods. Set to False to reduce latency (#​1428) (dd2f488)
• Add GeoSeries.boundary() (#​1435) (32cddfe)
• Add allow_large_results to peek (#​1448) (67487b9)
• Add groupby.rank() (#​1433) (3a633d5)
• Iloc multiple columns selection. (#​1437) (ddfd02a)
• Support interface for BigQuery managed functions (#​1373) (2bbf53f)
• Warn if default ingress_settings is used in remote_functions (#​1419) (dfd891a)

Bug Fixes

• Do not compare schema description during schema validation (#​1452) (03a3a56)
• Remove warnings for null index and partial ordering mode in prep for GA (#​1431) (6785aee)
• Warn if default cloud_function_service_account is used in remote_function (#​1424) (fe7463a)
• Window operations over JSON columns (#​1451) (0070e77)
• Write chunked text instead of dummy text for pdf chunk (#​1444) (96b0e8a)

Performance Improvements

• Speed up DataFrame corr, cov (#​1309) (c598c0a)

Documentation

• Add snippet for explaining the linear regression model prediction (#​1427) (7c37c7d)

v1.38.0

Compare Source

Features

• (Preview) Support diff aggregation for timestamp series. (#​1405) (abe48d6)
• Add GeoSeries.from_wkt() and GeoSeries.to_wkt() (#​1401) (2993b28)
• Support DF.array(copy=True) (#​1403) (693ed8c)
• Support routines with ARRAY return type in read_gbq_function (#​1412) (4b60049)

Bug Fixes

• Calling to_timdelta() over timedeltas no longer changes their values (#​1411) (650a190)
• Replace empty dict with None to avoid mutable default arguments (#​1416) (fa4e3ad)

Performance Improvements

• Avoid redundant SQL casts (#​1399) (6ee48d5)

Dependencies

• Remove scikit-learn and sqlalchemy as required dependencies (#​1296) (fd8bc89)

Documentation

• Add samples using SQL methods via the bigframes.bigquery module (#​1358) (f54e768)
• Add snippets for visualizing a time series and creating a time series model for the Limit forecasted values in time series model tutorial (#​1310) (c6c9120)

v1.37.0

Compare Source

Features

• (Preview) Support add, sub, mult, div, and more between timedeltas (#​1396) (ffa63d4)
• (Preview) Support comparison, ordering, and filtering for timedeltas (#​1387) (34d01b2)
• (Preview) Support subtraction in DATETIME/TIMESTAMP columns with timedelta columns (#​1390) (50ad3a5)
• JSON dtype support for read_pandas and Series constructor (#​1391) (44f4137)

Bug Fixes

• Ensure binops with pandas objects returns bigquery dataframes (#​1404) (3cee24b)

Performance Improvements

• Prune projections more aggressively (#​1398) (7990262)
• Simplify sum aggregate SQL text (#​1395) (0145656)
• Use simple null constraints to simplify queries (#​1381) (00611d4)

Documentation

• Add DataFrame.struct docs (#​1348) (7e9e93a)

v1.36.0

Compare Source

Features

• (Preview) Support addition between a timestamp and a timedelta (#​1369) (b598aa8)
• (Preview) Support casting floats and list-likes to timedelta series (#​1362) (65933b6)
• (Preview) Support timestamp subtractions (#​1346) (86b7e72)
• Add bigframes.bigquery.st_area and suggest it from GeoSeries.area (#​1318) (8b5ffa8)
• Add GeoSeries.from_xy() (#​1364) (3c3e14c)

Bug Fixes

• Dtype parameter ineffective in Series/DataFrame construction (#​1354) ([b9bdca8](https://redirect.github.com/googleapis/