下記テーブルのようなpd.DataFrame型（indexにunix時刻が入っている）があった時に，int型のtimestampをその後の処理上，datetime型に変更したいが，
pandasに用意されている関数をそのまま使うと，おかしな結果になってしまったので備忘録．
| x | y | z | |
|---|---|---|---|
| timestamp | |||
| 1450407402547 | -2.153091 | 1.582626 | 9.288803 |
| 1450407402577 | -2.368118 | 1.287659 | 9.227219 |
| 1450407402607 | -3.023911 | 4.160522 | 8.133606 |
| 1450407402637 | -2.316528 | 2.467163 | 9.581879 |
convert.py
df["time"] = pd.to_datetime(df.index , unit="ms")
df["time2"] =df.index
df["time3"] = df.time2.apply(lambda x: datetime.fromtimestamp(x/1000))
| x | y | z | time | time2 | time3 | |
|---|---|---|---|---|---|---|
| timestamp | ||||||
| 1450407402547 | -2.153091 | 1.582626 | 9.288803 | 2015-12-18 02:56:42.547 | 1450407402547 | 2015-12-18 11:56:42.546999 |
| 1450407402577 | -2.368118 | 1.287659 | 9.227219 | 2015-12-18 02:56:42.577 | 1450407402577 | 2015-12-18 11:56:42.576999 |
| 1450407402607 | -3.023911 | 4.160522 | 8.133606 | 2015-12-18 02:56:42.607 | 1450407402607 | 2015-12-18 11:56:42.607000 |
| 1450407402637 | -2.316528 | 2.467163 | 9.581879 | 2015-12-18 02:56:42.637 | 1450407402637 | 2015-12-18 11:56:42.637000 |
time2は下記のようにするとエラーが出たので無理やり作った．
df["time3"] = df.index.apply(lambda x: datetime.fromtimestamp(x/1000))
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-117-0c598f05317f> in <module>()
----> 1 df["time3"] = df.index.apply(lambda x: datetime.fromtimestamp(x/1000))
AttributeError: 'Int64Index' object has no attribute 'apply'
上記の結果からtime3のやり方だと正しくunix時刻が変換される．
pd.to_datetimeの仕様を確認せねば...