Better error reporting on bad datashapes

[mrocklin]

I have the following datashape

ds2 = """ var * {
  currently: {
    apparentTemperature: float64,
    dewPoint: float64,
    humidity: float64,
    icon: string,
    precipIntensity: int64,
    precipProbability: int64,
    pressure: float64,
    summary: string,
    temperature: float64,
    time: int64,
    visibility: int64,
    windBearing: int64,
    windSpeed: float64
    },
  daily: {
    data: 1 * {
      apparentTemperatureMax: float64,
      apparentTemperatureMaxTime: int64,
      apparentTemperatureMin: float64,
      apparentTemperatureMinTime: int64,
      cloudCover: int64,
      dewPoint: float64,
      humidity: float64,
      icon: string,
      moonPhase: float64,
      precipIntensity: float64,
      precipIntensityMax: float64,
      precipIntensityMaxTime: int64,
      precipProbability: float64,
      precipType: string,
      pressure: float64,
      summary: string,
      sunriseTime: int64,
      sunsetTime: int64,
      temperatureMax: float64,
      temperatureMaxTime: int64,
      temperatureMin: float64,
      temperatureMinTime: int64,
      time: int64,
      visibility: float64,
      windBearing: int64,
      windSpeed: float64
      }
    },
  flags: {isd_stations: 5 * string, sources: 1 * string, units: string},
  hourly: {
    data: 24 * {
      apparentTemperature: float64,
      cloudCover: ?int64,
      dewPoint: float64,
      humidity: float64,
      icon: string,
      precipIntensity: float64,
      precipProbability: float64,
      precipType: ?string,
      pressure: float64,
      summary: string,
      temperature: float64,
      time: int64,
      visibility: float64,
      windBearing: int64,
      windSpeed: float64
      },
    icon: string,
    summary: string
    },
  latitude: float64,
  longitude: float64,
  offset: int64,
  timezone: string
  }"""

I load data with this datashape and get an error:

from dynd import nd

with open('weather.json') as f:
    text = f.read()

>>> x = nd.parse_json(ds2, text.replace('isd-stations', 'isd_stations'))
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-25-4966077b3c8f> in <module>()
      4     text = f.read()
      5 
----> 6 x = nd.parse_json(ds2, text.replace('isd-stations', 'isd_stations'))

dynd/nd/array.pyx in dynd.nd.array.parse_json (/home/travis/build/libdynd/dynd-python/build/temp.linux-x86_64-3.4/array.cxx:8711)()

ValueError: parse error converting string "0.08" to int64

OK, so one of my int types need to be a float type. Which one?

[mrocklin]

Updated comment above

[izaid]

Okay, so I think there are a few things going on here.

1. I think Datashape and DyND are discovering slightly different types -- maybe just a mismatch in expectations between them, maybe a bug in one or both. I found the following substitutions were needed -- cloudCover: int64 -> float64, precipIntensityMaxTime: int64 -> ?int64, and precipType: string -> ?string. And there still might be other things...

2. This still is not enough to fix the dataset as ?string is not something DyND currently supports, but will (hopefully) soon.

[mwiebe]

In the json parser, the parsing code often catches errors like this and wraps them in a new exception including the parsing context of the line+column numbers. That's missing from the particular spot that is parsing this value.