230632 0 c -0.821511 1.103480 0 d NaN 0.917391 0 With Panel, describing the matching behavior is a bit more difficult, so the arithmetic methods instead (and perhaps confusingly?) give you the option to these methods generally assume that the indexes are sorted. They may be modified in the future to be a bit more flexible but as time series data is ordered most of the time anyway, this has not been a major 563010 0.332378 c 0.416574 -0.686906 d 1.140295 0.222904 Note that the following also works, but a bit less obvious / clean: In [114]: df.reindex(df.index - [’a’, ’d’]) Out[114]: one three two b -0.842389

226845 0 c -0.884702 -2.127985 0 d NaN 1.313212 0 With Panel, describing the matching behavior is a bit more difficult, so the arithmetic methods instead (and perhaps confusingly?) give you the option to these methods generally assume that the indexes are sorted. They may be modified in the future to be a bit more flexible but as time series data is ordered most of the time anyway, this has not been a major 069381 0.157464 c 0.076517 2.204503 d 0.973912 -0.339300 Note that the following also works, but a bit less obvious / clean: In [114]: df.reindex(df.index - [’a’, ’d’]) Out[114]: one three two b -0.476632

153404 0 c 0.582831 -1.261113 0 d NaN 0.747709 0 With Panel, describing the matching behavior is a bit more difficult, so the arithmetic methods instead (and perhaps confusingly?) give you the option to these methods generally assume that the indexes are sorted. They may be modified in the future to be a bit more flexible but as time series data is ordered most of the time anyway, this has not been a major 304932 0.151528 c -0.926851 0.334262 d 0.230059 -0.517650 Note that the following also works, but a bit less obvious / clean: In [114]: df.reindex(df.index - [’a’, ’d’]) Out[114]: one three two b -0.133357

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1673 34.9.1.21 pandas.MultiIndex.lexsort_depth . . . . . . . . . . . . . . . . . . . . . . . . . . . 1673 34.9.1.22 pandas.MultiIndex.name . . . to unsigned 64-bit integers (GH4471, GH14982) 14 Chapter 1. What’s New pandas: powerful Python data analysis toolkit, Release 0.20.3 • Bug in Series.unique() in which unsigned 64-bit integers were causing DataFrame construction in which unsigned 64-bit integer elements were being converted to objects (GH14881) • Bug in pd.read_csv() in which unsigned 64-bit integer elements were being improperly converted

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1772 34.10.1.21pandas.MultiIndex.lexsort_depth . . . . . . . . . . . . . . . . . . . . . . . . . . . 1772 xxxiii 34.10.1.22pandas.MultiIndex.name unsigned 64-bit integers (GH4471, GH14982) • Bug in Series.unique() in which unsigned 64-bit integers were causing overflow (GH14721) • Bug in DataFrame construction in which unsigned 64-bit integer elements pd.read_csv() in which unsigned 64-bit integer elements were being improperly converted to the wrong data types (GH14983) • Bug in pd.unique() in which unsigned 64-bit integers were causing overflow (GH14915)

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1619 pandas.MultiIndex.lexsort_depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1619 pandas.MultiIndex.name . . Series rather than SparseSeries (GH13999) Indexer dtype changes Note: This change only affects 64 bit python running on Windows, and only affects relatively advanced indexing operations Methods such as that can hold a pointer (GH3033, GH13972). These types are the same on many platform, but for 64 bit python on Windows, np.int_ is 32 bits, and np.intp is 64 bits. Changing this behavior improves performance

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1624 pandas.MultiIndex.lexsort_depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1624 pandas.MultiIndex.name . . Series rather than SparseSeries (GH13999) Indexer dtype changes Note: This change only affects 64 bit python running on Windows, and only affects relatively advanced indexing operations Methods such as Python data analysis toolkit, Release 0.19.1 These types are the same on many platform, but for 64 bit python on Windows, np.int_ is 32 bits, and np.intp is 64 bits. Changing this behavior improves performance

metacharacters were being treated as regexs even when regex=False (GH6777). • Bug in timedelta ops on 32-bit platforms (GH6808) • Bug in setting a tz-aware index directly via .index (GH6785) • Bug in expressions functions that used *args‘‘ or **kwargs and returned an empty result (GH6952) • Bug in sum/mean on 32-bit platforms on overflows (GH6915) • Moved Panel.shift to NDFrame.slice_shift and fixed to respect multiple Out[23]: a int64 dtype: object Keep in mind that DataFrame(np.array([1,2])) WILL result in int32 on 32-bit platforms! Upcasting Gotchas Performing indexing operations on integer type data can easily upcast

unsigned 64-bit integers (GH4471, GH14982) • Bug in Series.unique() in which unsigned 64-bit integers were causing overflow (GH14721) • Bug in DataFrame construction in which unsigned 64-bit integer elements pd.read_csv() in which unsigned 64-bit integer elements were being improperly converted to the wrong data types (GH14983) • Bug in pd.unique() in which unsigned 64-bit integers were causing overflow (GH14915) (GH14915) • Bug in pd.value_counts() in which unsigned 64-bit integers were being erroneously truncated in the output (GH14934) 1.2.1.8 GroupBy on Categoricals In previous versions, .groupby(..., sort=False)

Performance Improvements • Bug Fixes 1.1.1 API changes • Indexing in MultiIndex beyond lex-sort depth is now supported, though a lexically sorted index will have a better performance. (GH2646) In [1]: Out[2]: jolie jim joe 0 x 0.179356 x 0.908835 1 z 0.571981 y 0.851401 In [3]: df.index.lexsort_depth Out[3]: 1 # in prior versions this would raise a KeyError # will now show a PerformanceWarning In Out[6]: jolie jim joe 0 x 0.179356 x 0.908835 1 y 0.851401 z 0.571981 In [7]: df2.index.lexsort_depth Out[7]: 2 In [8]: df2.loc[(1,’z’)] Out[8]: jolie jim joe 1 z 0.571981 • Bug in unique of Series