How to create a new column from other column by condition

What about np.where:

df["action_price"] = np.where(df.buy_action == 1, df.enterhigh, df.exithigh)

It takes values from enter_high if buy_action is 1, else it takes from exit_high.

And we get

                date  enterhigh  exithigh  buy_action  action_price
2017-08-20  06:00:00    4149.99   4044.00         1.0       4149.99
2017-08-20  23:00:00    4211.08   4068.50        -1.0       4068.50
2017-08-22  17:00:00    4041.23   3750.08         1.0       4041.23
2017-08-23  19:00:00    4265.80   4103.51        -1.0       4103.51
2017-08-24  21:00:00    4250.94   4136.33         1.0       4250.94

Note that any value of buy_action that is not 1 (e.g. -1, 2) will yield values from exithigh.

That's a good use case for the apply method, that'll take a row and returns a new value :

df["action_price"] = df.apply(lambda x: x.enterhigh if x.buy_action == 1 else x.exithigh, axis=1)

You could use a list comprehension:

df['new column'] = [df.loc[i, 'exithigh'] for i in df.index if df.loc[i, 'buy_action']>0 else -1]

Using numpy select:

df['action_price'] = np.select([df.buy_action.eq(1)], [df.enterhigh], default=df.exithigh)