wrong S2 cell ID for given latitude and longitude

[adrianomitre]

(main)> require 'poke-api'
=> true
(main)> lat = -23.59260639636282
=> -23.59260639636282
(main)> lng = -46.680191904206524
=> -46.680191904206524
(main)> Poke::API::Helpers.get_cells(lat, lng)
=> [
    [ 0] 10722603741376151552,
    [ 1] 10722603743523635200,
    [ 2] 10722603745671118848,
    [ 3] 10722603747818602496,
    [ 4] 10722603749966086144,
    [ 5] 10722603752113569792,
    [ 6] 10722603754261053440,
    [ 7] 10722603756408537088,
    [ 8] 10722603758556020736,
    [ 9] 10722603760703504384,
    [10] 10722603762850988032,
    [11] 10722603764998471680,
    [12] 10722603767145955328,
    [13] 10722603769293438976,
    [14] 10722603771440922624,
    [15] 10722603773588406272,
    [16] 10722603775735889920,
    [17] 10722603777883373568,
    [18] 10722603780030857216,
    [19] 10722603782178340864,
    [20] 10722603784325824512
]

which does not contain the S2 Cell ID 10722603763924729856, whose center is lat, lng.

Contrast this with sidewalklabs/s2sphere: Python implementation of the S2 geometry library:

>>> import s2sphere
>>> lat = -23.59260639636282
>>> lng = -46.680191904206524
>>> cell = s2sphere.Cell.from_lat_lng(s2sphere.LatLng.from_degrees(lat, lng))
>>> cell.id().parent(12).id()
10722603763924729856

[nabeelamjad]

Good find, unfortunately I don't think I've got the time now to debug this as this project has been deprecated since almost a year ago and I'm quite busy these days.

The S2 implementation was a quick conversion of some functionality to make it work for calculating cells in use with Pokemon Go, during the conversion it appeared to work fine (as in I had values that were the same as s2sphere from Python, granted there's no test suite or anything for this project).

I'd probably refer to people using an actual port or native binding of Google's S2 Geometry for the full functionality, though I don't think one exists for Ruby (which is why I had to throw something together).

[adrianomitre]

So maybe a warning on related source files and/or README would suffice... Em seg, 17 de jul de 2017 às 21:56, nabeelamjad <notifications@github.com> escreveu:

Good find, unfortunately I don't think I've got the time now to debug this as this project has been deprecated since almost one year ago and I'm quite busy these days. The S2 implementation was a quick conversion of some functionality to make it work for calculating cells in use with Pokemon Go, during the conversion it appeared to work fine (as in I had values that were the same as s2sphere from Python, granted there's no test suite or anything for this project). I'd probably refer to people using an actual port or native binding of Google's S2 Geometry for the full functionality, though I don't think one exists for Ruby. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#48 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAb4ZFWg6uovdUDymgEYts-O_IV0uO5Qks5sPAKzgaJpZM4OaqaR> .

-- *Adriano Mitre*