Try CSV if binary read fails

[rweigel]

A common error is that the length for a string in the /info response is wrong. In this case, the binary read will fail, but the csv read will work.

If the binary read fails, request csv and if it works, warn that there is something wrong with the server's binary response.

[Tiger767]

I find a similar problem to be found in this code:

from hapiclient import hapi
server = 'https://cdaweb.gsfc.nasa.gov/hapi'
options = {'logging': False, 'usecache': True, 'cachedir': './hapicache'}
current_time = dt.datetime(year=2018, month=1, day=1)
fgm_data, meta = hapi(server, 'MMS1_FGM_SRVY_L2@0', 'mms1_fgm_b_gsm_srvy_l2_clean',
     current_time.strftime("%Y-%m-%d") + 'T00:00:00Z',
     (current_time + dt.timedelta(days=1)).strftime("%Y-%m-%d") + 'T00:00:00Z', **options)

The code above result:

array([(b'2018-01-01T00:00:00.0598', [ 4.90015926e-310,  1.04347723e-309,  3.48698000e-310,  1.04372585e-309]),
       (b'\x00\x80\x9c\xa83@2018-01-01T00:00:0', [ 6.52398915e-038, -2.68156249e+154, -1.49166921e-154,  2.00000048e+000]),
       (b']B"\xc0\x00\x00\x00@\xe7\x833@2018-01-01T0', [ 3.25510386e-086,  1.37100058e-071, -6.22868538e-006,  8.20510554e-163]),
       ...,
       (b'\x8dI\x12@\x00\x00\x00\xa0\xf3\xec\xf4?\x00\x00\x00`\xd9>\x13@2018', [ 2.99587260e-066,  5.06871946e-086,  1.52005163e-047,  5.76195421e+129]),
       (b'\xe7\xbf\x00\x00\x00\x80\xafG\x12@\x00\x00\x00\xc0\x06\xdc\xf4?\x00\x00\x00@p;', [ 1.26172266e-076,  2.42295120e-028,  5.87796863e-062,  3.24922494e+126]),
       (b'\x00\x00\x00`t{\xe7\xbf\x00\x00\x00\xa0\xdaI\x12@\x00\x00\x00\xc0k\xae\xf4?', [ 4.80765724e+000,  5.27365195e-091,  4.08823769e-033,  8.16645805e-043])],
      dtype=[('Time', 'S24'), ('mms1_fgm_b_gsm_srvy_l2_clean', '<f8', (4,))])

Solution as noted by @rweigel works: Add 'format': 'csv' to options dictionary.

As also noted by @rweigel problem arises from incorrect length in time parameter meta data that was obtained by the /info request

[rweigel]

One bug this highlights is the fact that the binary reader did not throw an error. The binary reader needs to check the response length against what is expected from the lengths, sizes, and types in the metadata.