large tolerances in test_pvsystem.test_singlediode_floats_ivcurve

[mikofski]

the absolute tolerance to compare test and expected output is 3 in test_pvsystem.test_singlediode_floats_ivcurve which allows almost any values to pass:

e.g.:

>>> out['i_sc'], expected['i_sc']
(6.9651723221583133, 6.96467476139632)
>>> np.isclose(out['i_sc'], expected['i_sc'])
False
>>> np.isclose(out['i_sc'], expected['i_sc'], atol=3)
True
>>> out['v_oc'], expected['v_oc']
(8.1063001465863103, 8.106300146586308)
>>> out['v_mp'], expected['v_mp']
(6.2243393757884284, 6.221535886625464)
>>> out['i_mp'], expected['i_mp']
(6.1362673597376753, 6.13902517979357)
>>> out['i'], expected['i']
(array([  7.00000000e+00,   6.53439455e+00,   3.48054918e-14]),
 array([  6.96517200e+00,   6.75588200e+00,   2.57571700e-14]))
>>> out['v'], expected['v']
(array([-0.7       ,  5.5749351 ,  8.10630015]),
 array([ 0.     ,  4.05315,  8.1063 ]))

IMHO these tolerance should be closer to 1e-3 maybe, or maybe set rtol?

[adriesse]

Looks like a serious typo! I vote for rtol.

[wholmgren]

Yes. Probably best to merge a fix for this before merging #409. Do you want to make the PR? I can do a quick one this afternoon if you want to stay focused on your current pvlib work.

[markcampanelli]

Nice catch!