Inc. 3 Confidential │ ©2021 VMware, Inc. Brin Index 在Greenplum 7中的 理论与实践 陈金豹，VMWare内核工程师 Confidential │ ©2021 VMware, Inc. 目录 Brin Index On Heap AppendOnly Table Brin在AppendOnly Table上的实现 Table上的实现 性能测试 5 Confidential │ ©2021 VMware, Inc. Block Range Index 存储数据块中元组字段的最 大最小值，用于过滤不符合条 件的数据块 1 3 2 5 7 8 8 10 9 11 11 12 [1, 5] [7, 10] [9, 12] Brin Heap Confidential │ ©2021 VMware 1-((B-1)/B)^(N*a) = 1- 0.000045 Confidential │ ©2021 VMware, Inc. Brin的使用场景 表非常大 数据有一定的分布特性 我们不想再index上付出太多存储空间 Confidential │ ©2021 VMware, Inc. Brin Scan select * from t where a > 1 and a < 8;

. . . . . . . . . . . 444 13.18 Index objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444 13.19 Set / Reset Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453 14.2 Advanced indexing with hierarchical index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459 14.3 The need for sortedness with . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471 14.6 Float64Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474

. . . . 22 i 1.3.2.2 Map on Index types now return other Index types . . . . . . . . . . . . . . . . . . 23 1.3.2.3 Accessing datetime fields of Index now return Index . . . . . . . . . . . . . . . . . . . . . . . . . 27 1.3.2.9 Memory Usage for Index is more Accurate . . . . . . . . . . . . . . . . . . . . . . 27 1.3.2.10 DataFrame.sort_index changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 1.3.2.14 Index.intersection and inner join now preserve the order of the left Index . . . . . . 32 1.3.2.15 Pivot Table always returns a DataFrame

when attempting to create columns . . . . . . . . . . . . . . . 9 1.2.1.4 drop now also accepts index/columns keywords . . . . . . . . . . . . . . . . . . . 10 1.2.1.5 rename, reindex now also accept . . . . . . . 17 1.2.2.5 Iteration of Series/Index will now return Python scalars . . . . . . . . . . . . . . . 17 1.2.2.6 Indexing with a Boolean Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 1.5.2.2 Map on Index types now return other Index types . . . . . . . . . . . . . . . . . . 52 1.5.2.3 Accessing datetime fields of Index now return Index . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 9.10 Sorting by index and value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 9.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 10.19 Float64Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292 10.22 Index objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

casefold() method to removes all case distinctions present in a string (GH25405) • DataFrame.set_index() now works for instances of abc.Iterator, provided their output is of the same length as the calling behavior of the sort parameter matches that of Index.union() (GH24994) • RangeIndex.union() now supports the sort argument. If sort=False an unsorted Int64Index is always returned. sort=None is the default default and returns a monotonically increasing RangeIndex if possible or a sorted Int64Index if not (GH24471) • TimedeltaIndex.intersection() now also supports the sort keyword (GH24471) • DataFrame.rename()

casefold() method to removes all case distinctions present in a string (GH25405) • DataFrame.set_index() now works for instances of abc.Iterator, provided their output is of the same length as the calling behavior of the sort parameter matches that of Index.union() (GH24994) • RangeIndex.union() now supports the sort argument. If sort=False an unsorted Int64Index is always returned. sort=None is the default default and returns a monotonically increasing RangeIndex if possible or a sorted Int64Index if not (GH24471) • TimedeltaIndex.intersection() now also supports the sort keyword (GH24471) • DataFrame.rename()

creating a markdown table (GH11052) In [1]: df = pd.DataFrame({"A": [1, 2, 3], "B": [1, 2, 3]}, index=['a', 'a', 'b']) In [2]: print(df.to_markdown()) | | A | B | |:---|----:|----:| | a | 1 | 1 | | control when wide columns are trun- cated (GH9784) • Added the na_value argument to Series.to_numpy(), Index.to_numpy() and DataFrame. to_numpy() to control the value used for missing data (GH30322) • MultiIndex have gained ignore_index keyword to be able to reset index after sorting (GH30114) • DataFrame.sort_index() and Series.sort_index() have gained ignore_index keyword to reset index (GH30114) • DataFrame

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 9.10 Sorting by index and value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 9.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 10.19 Float64Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265 10.22 Index objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Semi-Month Offsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 New Index methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Google BigQuery . 26 Index + / - no longer used for set operations . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Index.difference and .symmetric_difference changes . . . . . . . . . . . . . 29 Index.unique unique consistently returns Index . . . . . . . . . . . . . . . . . . . . . . . . . 29 MultiIndex constructors, groupby and set_index preserve categorical dtypes . . . . 30 read_csv will progressively enumerate