`interp` returns strange value when `x` is outside the grid range

[shizejin]

>>> from interpolation import interp
>>> import numpy as np
>>> x = np.array([0., 1.])
>>> y = np.array([1., 2.])
>>> interp(x, y, 0.5)
1.5
>>> interp(x, y, 1.5)
-1.5555469624081107e+231
>>> interp(x, y, 2.)
-3.1110939248162215e+231

[albop]

Yeah, that's a lot...

Le dim. 23 sept. 2018 à 06:46, Zejin Shi notifications@github.com a écrit :

from interpolation import interp>>> import numpy as np>>> x = np.array([0., 1.])>>> y = np.array([1., 2.])>>> interp(x, y, 0.5)1.5>>> interp(x, y, 1.5)-1.5555469624081107e+231>>> interp(x, y, 2.)-3.1110939248162215e+231

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/EconForge/interpolation.py/issues/29, or mute the thread https://github.com/notifications/unsubscribe-auth/AAQ5KddLnp_yx7jCBeL0is3rcMFANFalks5udxJBgaJpZM4W1js1 .

[jstac]

@albop Just FYI, @shizejin is helping to hook up the interpolations library to the QuantEcon lectures.

[albop]

@shizejin and @jstac Cool! @shizejin: I think I fixed it in the PR #30. Here is what was happening:

• there is a function get_index which returns (i, λ) where i is the index of the cell to interpolate and λ a barycentric coordinate in that cell
• i was not restricted to lie within the acceptable range which contains only one value in your example (0,) so there was an out of bounds error and a random picked value was picked. For some reason on my machine it is a finite one so I didn't notice. Wasn't so finite on yours.
• the fix just clamps the value of i inside the admissible range (see i = clamp(i, 0, N-2) in https://github.com/EconForge/interpolation.py/pull/30/commits/1294d63e4b035d3b1a7a29e493c2aa83cd255289) . When extrapolation happens then the barycentric is not contained in [0,1] anymore, but that is ok.
• makes me wonder whether we should activate out of bounds errors when running the tests...

[shizejin]

Thank you for your quick response @albop ! interp is now working correctly in the lecture notebook.