ValueError: cannot compute fingerprint of empty list

[dickreuter]

from numba import jit
@jit()
def create_card_deck():
    values = "23456789TJQKA"
    suites = "CDHS"
    Deck = []
    for x in values:
        for y in suites:
            Deck.append(x + y)
    return Deck

create_card_deck()

[synapticarbors]

A work around is to create the list with an initial item of the same type as you are going to add and then throw it away before you return:

from numba import jit

@jit
def create_card_deck():
    values = "23456789TJQKA"
    suites = "CDHS"
    Deck = ['x']
    for x in values:
        for y in suites:
            Deck.append(x + y)
    return Deck[1:]

[dickreuter]

Ok makes sense. Thanks

On Tuesday, 18 October 2016, Joshua Adelman notifications@github.com
wrote:

A work around is to create the list with an initial item of the same type
as you are going to add and then throw it away before you return:

from numba import jit
@jitdef create_card_deck():
values = "23456789TJQKA"
suites = "CDHS"
Deck = ['x']
for x in values:
for y in suites:
Deck.append(x + y)
return Deck[1:]

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2152 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/ABMrfkyaD62cIBy3_sLfuMMk--MtHiGzks5q1NQogaJpZM4KX9jG
.

[stuartarchibald]

Related comments on what needs to happen to fix this are at the bottom of #2184

[Permafacture]

So, how does the closing of #2184 affect this issue? Is it now intentional to disallow empty lists being passed into nopython jitted functions?

[stuartarchibald]

As of 0.42.0 the original sample works (updates have been made to type inference to help type empty lists). Closing.

[Mostro-Complexity]

When I use version 0.44.1 numba, the above problem still occurs.
my code:

def func(X, a, b):
    if len(a) > 0 and len(b) > 0:
        for i in range(X.shape[1]):
            X[:, :i] += i

    return np.empty(X.shape[0], dtype=np.float64)


func(np.zeros((10, 10)), [], [])

Error message:

Exception has occurred: ValueError
cannot compute fingerprint of empty list

[stuartarchibald]

@Mostro-Complexity reflected lists (what you are using) are in the process of being deprecated, please take a look at their replacement, numba.typed.List, docs are here. Thanks.

[shaunc]

I'm getting the equivalent error for set. There doesn't seem to be any equivalent to numba.typed.List for set? (At least, not in doc.) Should I open a tracking issue ... or is this otherwise in the works?

Update:

Ah -- https://numba.pydata.org/numba-doc/latest/reference/deprecation.html#expected-replacement

Upshot: typed set is at least documented as not yet implemented.

[esc]

Upshot: typed set is at least documented as not yet implemented.

Yup it's on the wishlist. It is quite the undertaking and will require lot's of code to be written.It will however be reasonably well isolated from the rest of the code base, since the feature will be largely additive. And, in addition there are two existing typed container implementations, List and Dict that can be cribbed from. Essentially, it involves taking the set implementation from cpython and transplanting it into Numba (and it's runtime) and making it usable from there.