添加到HDFStore时,分类数据已损坏

时间:2015-10-26 10:13:00

标签: python pandas hdf5

Pandas 0.16.2使"11-9553АВП"类别为空字符串,0.17使'utf-8' codec can't decode byte 0xd0

store = pd.HDFStore(r"D:\test10.h5")
store["data"] = pd.Series(("cat", "cat", "11-9553АВП", "hippo")).astype('category')
print(store["data"])
store.close()

调试后,我看到pytables _read_records截断了某些类别的1个字节。或者也许在写入HDF元字段时它们会被截断。

UPD:好的。 https://github.com/pydata/pandas/issues/11234已修复https://github.com/pydata/pandas/commit/26db1729f2947d95d9a2e51878ba0f297fe24165

0 个答案:

没有答案