python "in" function works differently for lists and numpy arrays

Question

If I have a list of tuples, the "in" function tells me exactly whether the tuple is in the list.

>>> a = [(1,2),(3,4)]

>>> (1,2) in a
True

>>> (1,3) in a
False

That's what I want. But when I do the same thing for b = np.array(a), both give True:

>>> b = np.array(a)

>>> (1,2) in b
True

>>> (1,3) in b
True

So it seems as though the in function for numpy arrays does not check the tuple anymore, but the single values.

Why is this?
How can I achive a truth request regarding whether a 2d element is in a numpy array as in the example above?

Operators work differently for *all* types. You would not expect ``+`` to work the same for integers, lists and strings, would you? — MisterMiyagi, Dec 23 '21 at 10:02
[Related](https://stackoverflow.com/questions/18320624/how-does-contains-work-for-ndarrays) — hilberts_drinking_problem, Dec 23 '21 at 10:05
Does this answer your question? ["In" operator for numpy arrays?](https://stackoverflow.com/questions/39452843/in-operator-for-numpy-arrays) — Dariusz Krynicki, Dec 23 '21 at 10:11
Thanks! So it's clear it doesn't work for arrays as I need it. But what's the alternative for arrays? Looping through the single elements would take too much time, I suppose, would it not? I need to handle a lot of tuples (3.6e6 times 10) — Interferenzbrille _, Dec 23 '21 at 10:13
Of course, **there are no `tuple` objects in your array**. Why do you *expect* this to work at all?? — juanpa.arrivillaga, Dec 23 '21 at 10:27
@DariuszKrynicki thanks you. np.isin doesn't help, it does the same as in, but the first answer is probably the solution. I don't understand it, though, but I have to try, then. Thanks — Interferenzbrille _, Dec 23 '21 at 10:30
@juanpa.arrivillaga why are there no tuple objects in my array anymore? — Interferenzbrille _, Dec 23 '21 at 10:30
@Interferenzbrille_ why do you think there should be any? What would be the point of creating the array? Numpy arrays are wrappers over primitive arrays. They do not contain *python objects at all* (unless you specify dtype=object, but a dtype=object array is just a less performant python list) — juanpa.arrivillaga, Dec 23 '21 at 10:35
To efficiently check if a "row" is in an array, then probably you can convert to a flattened structured dtype, then convert the tuples to scalars of that same dtype — juanpa.arrivillaga, Dec 23 '21 at 10:40
You could try to use pure Python set difference to get decent performance in this case. For example, if `data = set([(1,2), (1,3)])` and `query = set([(1,2),(1,3)])`, you could compute `query - data` fairly fast (.4 seconds for 3 million 10-tuples in my case). — hilberts_drinking_problem, Dec 23 '21 at 10:41
@juanpa.arrivillaga I thought they would contain stuff, but yes, makes sense that they don't. Thanks! — Interferenzbrille _, Dec 23 '21 at 10:52
(1,3) in np.array(a).tolist() gives False, (1,2) in np.array(a).tolist() does as well, though. Thanks yall! I concluded, for what I wanted to do, I just need lists. — Interferenzbrille _, Dec 28 '21 at 17:38

python "in" function works differently for lists and numpy arrays

0 Answers0