|
| 1 | +from datetime import datetime, date, timedelta |
| 2 | +from typing import List, Tuple, Union, Iterable |
| 3 | + |
| 4 | +import pandas as pd |
| 5 | + |
| 6 | +from delphi_epidata import Epidata |
| 7 | + |
| 8 | +def date_generator(startdate, enddate): |
| 9 | + while startdate <= enddate: |
| 10 | + yield startdate.strftime('%Y-%m-%d') |
| 11 | + startdate = startdate + timedelta(days=1) |
| 12 | + |
| 13 | + |
| 14 | + |
| 15 | +def metadata(): |
| 16 | + response = Epidata._request("covidcast_meta") |
| 17 | + |
| 18 | + if response["result"] != 1: |
| 19 | + # Something failed in the API and we did not get real metadata |
| 20 | + raise RuntimeError("Error when fetching metadata from the API", |
| 21 | + response["message"]) |
| 22 | + |
| 23 | + df = pd.DataFrame.from_dict(response["epidata"]) |
| 24 | + return df |
| 25 | + |
| 26 | + |
| 27 | +def signal( |
| 28 | + data_source: str, |
| 29 | + signal: str, # pylint: disable=W0621 |
| 30 | + start_day: date = None, |
| 31 | + end_day: date = None, |
| 32 | + geo_type: str = "county", |
| 33 | + geo_values: Union[str, Iterable[str]] = "*", |
| 34 | + as_of: date = None, |
| 35 | + issues: Union[date, Tuple[date], List[date]] = None, |
| 36 | + lag: int = None, |
| 37 | + time_type: str = "day", |
| 38 | +) -> Union[pd.DataFrame, None]: |
| 39 | + """Download a Pandas data frame for one signal. |
| 40 | +
|
| 41 | + Obtains data for selected date ranges for all geographic regions of the |
| 42 | + United States. Available data sources and signals are documented in the |
| 43 | + `COVIDcast signal documentation |
| 44 | + <https://cmu-delphi.github.io/delphi-epidata/api/covidcast_signals.html>`_. |
| 45 | + Most (but not all) data sources are available at the county level, but the |
| 46 | + API can also return data aggregated to metropolitan statistical areas, |
| 47 | + hospital referral regions, or states, as desired, by using the ``geo_type`` |
| 48 | + argument. |
| 49 | +
|
| 50 | + The COVIDcast API tracks updates and changes to its underlying data, and |
| 51 | + records the first date each observation became available. For example, a |
| 52 | + data source may report its estimate for a specific state on June 3rd on June |
| 53 | + 5th, once records become available. This data is considered "issued" on June |
| 54 | + 5th. Later, the data source may update its estimate for June 3rd based on |
| 55 | + revised data, creating a new issue on June 8th. By default, ``signal()`` |
| 56 | + returns the most recent issue available for every observation. The |
| 57 | + ``as_of``, ``issues``, and ``lag`` parameters allow the user to select |
| 58 | + specific issues instead, or to see all updates to observations. These |
| 59 | + options are mutually exclusive; if you specify more than one, ``as_of`` will |
| 60 | + take priority over ``issues``, which will take priority over ``lag``. |
| 61 | +
|
| 62 | + Note that the API only tracks the initial value of an estimate and *changes* |
| 63 | + to that value. If a value was first issued on June 5th and never updated, |
| 64 | + asking for data issued on June 6th (using ``issues`` or ``lag``) would *not* |
| 65 | + return that value, though asking for data ``as_of`` June 6th would. |
| 66 | +
|
| 67 | + Note also that the API enforces a maximum result row limit; results beyond |
| 68 | + the maximum limit are truncated. This limit is sufficient to fetch |
| 69 | + observations in all counties in the United States on one day. This client |
| 70 | + automatically splits queries for multiple days across multiple API calls. |
| 71 | + However, if data for one day has been issued many times, using the |
| 72 | + ``issues`` argument may return more results than the query limit. A warning |
| 73 | + will be issued in this case. To see all results, split your query across |
| 74 | + multiple calls with different ``issues`` arguments. |
| 75 | +
|
| 76 | + See the `COVIDcast API documentation |
| 77 | + <https://cmu-delphi.github.io/delphi-epidata/api/covidcast.html>`_ for more |
| 78 | + information on available geography types, signals, and data formats, and |
| 79 | + further discussion of issue dates and data versioning. |
| 80 | +
|
| 81 | + :param data_source: String identifying the data source to query, such as |
| 82 | + ``"fb-survey"``. |
| 83 | + :param signal: String identifying the signal from that source to query, |
| 84 | + such as ``"smoothed_cli"``. |
| 85 | + :param start_day: Query data beginning on this date. Provided as a |
| 86 | + ``datetime.date`` object. If ``start_day`` is ``None``, defaults to the |
| 87 | + first day data is available for this signal. If ``time_type == "week"``, then |
| 88 | + this is rounded to the epiweek containing the day (i.e. the previous Sunday). |
| 89 | + :param end_day: Query data up to this date, inclusive. Provided as a |
| 90 | + ``datetime.date`` object. If ``end_day`` is ``None``, defaults to the most |
| 91 | + recent day data is available for this signal. If ``time_type == "week"``, then |
| 92 | + this is rounded to the epiweek containing the day (i.e. the previous Sunday). |
| 93 | + :param geo_type: The geography type for which to request this data, such as |
| 94 | + ``"county"`` or ``"state"``. Available types are described in the |
| 95 | + COVIDcast signal documentation. Defaults to ``"county"``. |
| 96 | + :param geo_values: The geographies to fetch data for. The default, ``"*"``, |
| 97 | + fetches all geographies. To fetch one geography, specify its ID as a |
| 98 | + string; multiple geographies can be provided as an iterable (list, tuple, |
| 99 | + ...) of strings. |
| 100 | + :param as_of: Fetch only data that was available on or before this date, |
| 101 | + provided as a ``datetime.date`` object. If ``None``, the default, return |
| 102 | + the most recent available data. If ``time_type == "week"``, then |
| 103 | + this is rounded to the epiweek containing the day (i.e. the previous Sunday). |
| 104 | + :param issues: Fetch only data that was published or updated ("issued") on |
| 105 | + these dates. Provided as either a single ``datetime.date`` object, |
| 106 | + indicating a single date to fetch data issued on, or a tuple or list |
| 107 | + specifying (start, end) dates. In this case, return all data issued in |
| 108 | + this range. There may be multiple rows for each observation, indicating |
| 109 | + several updates to its value. If ``None``, the default, return the most |
| 110 | + recently issued data. If ``time_type == "week"``, then these are rounded to |
| 111 | + the epiweek containing the day (i.e. the previous Sunday). |
| 112 | + :param lag: Integer. If, for example, ``lag=3``, fetch only data that was |
| 113 | + published or updated exactly 3 days after the date. For example, a row |
| 114 | + with ``time_value`` of June 3 will only be included in the results if its |
| 115 | + data was issued or updated on June 6. If ``None``, the default, return the |
| 116 | + most recently issued data regardless of its lag. |
| 117 | + :param time_type: The temporal resolution to request this data. Most signals |
| 118 | + are available at the "day" resolution (the default); some are only |
| 119 | + available at the "week" resolution, representing an MMWR week ("epiweek"). |
| 120 | + :returns: A Pandas data frame with matching data, or ``None`` if no data is |
| 121 | + returned. Each row is one observation on one day in one geographic location. |
| 122 | + Contains the following columns: |
| 123 | +
|
| 124 | + ``geo_value`` |
| 125 | + Identifies the location, such as a state name or county FIPS code. The |
| 126 | + geographic coding used by COVIDcast is described in the `API |
| 127 | + documentation here |
| 128 | + <https://cmu-delphi.github.io/delphi-epidata/api/covidcast_geography.html>`_. |
| 129 | +
|
| 130 | + ``signal`` |
| 131 | + Name of the signal, same as the value of the ``signal`` input argument. Used for |
| 132 | + downstream functions to recognize where this signal is from. |
| 133 | +
|
| 134 | + ``time_value`` |
| 135 | + Contains a `pandas Timestamp object |
| 136 | + <https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Timestamp.html>`_ |
| 137 | + identifying the date this estimate is for. For data with ``time_type = "week"``, this |
| 138 | + is the first day of the corresponding epiweek. |
| 139 | +
|
| 140 | + ``issue`` |
| 141 | + Contains a `pandas Timestamp object |
| 142 | + <https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Timestamp.html>`_ |
| 143 | + identifying the date this estimate was issued. For example, an estimate |
| 144 | + with a ``time_value`` of June 3 might have been issued on June 5, after |
| 145 | + the data for June 3rd was collected and ingested into the API. |
| 146 | +
|
| 147 | + ``lag`` |
| 148 | + Integer giving the difference between ``issue`` and ``time_value``, |
| 149 | + in days. |
| 150 | +
|
| 151 | + ``value`` |
| 152 | + The signal quantity requested. For example, in a query for the |
| 153 | + ``confirmed_cumulative_num`` signal from the ``usa-facts`` source, |
| 154 | + this would be the cumulative number of confirmed cases in the area, as |
| 155 | + of the ``time_value``. |
| 156 | +
|
| 157 | + ``stderr`` |
| 158 | + The value's standard error, if available. |
| 159 | +
|
| 160 | + ``sample_size`` |
| 161 | + Indicates the sample size available in that geography on that day; |
| 162 | + sample size may not be available for all signals, due to privacy or |
| 163 | + other constraints. |
| 164 | +
|
| 165 | + ``geo_type`` |
| 166 | + Geography type for the signal, same as the value of the ``geo_type`` input argument. |
| 167 | + Used for downstream functions to parse ``geo_value`` correctly |
| 168 | +
|
| 169 | + ``data_source`` |
| 170 | + Name of the signal source, same as the value of the ``data_source`` input argument. Used for |
| 171 | + downstream functions to recognize where this signal is from. |
| 172 | +
|
| 173 | + Consult the `signal documentation |
| 174 | + <https://cmu-delphi.github.io/delphi-epidata/api/covidcast_signals.html>`_ |
| 175 | + for more details on how values and standard errors are calculated for |
| 176 | + specific signals. |
| 177 | +
|
| 178 | + """ |
| 179 | + if start_day > end_day: |
| 180 | + raise ValueError( |
| 181 | + "end_day must be on or after start_day, but " f"start_day = '{start_day}', end_day = '{end_day}'" |
| 182 | + ) |
| 183 | + |
| 184 | + time_values = list(date_generator(start_day, end_day)) |
| 185 | + issues = list(date_generator(start_day, end_day)) #TODO placesholder |
| 186 | + response = Epidata.covidcast(data_source, signal, time_type=time_type, |
| 187 | + geo_type=geo_type, time_values=time_values, |
| 188 | + geo_value=geo_values, as_of=as_of, |
| 189 | + issues=issues, lag=lag) |
| 190 | + if response["result"] != 1: |
| 191 | + # Something failed in the API and we did not get real metadata |
| 192 | + raise RuntimeError("Error when fetching metadata from the API", |
| 193 | + response["message"]) |
| 194 | + |
| 195 | + api_df = pd.DataFrame.from_dict(response["epidata"]) |
| 196 | + api_df["issue"] = pd.to_datetime(api_df["issue"], format='%Y%m%d') |
| 197 | + api_df["time_value"] = pd.to_datetime(api_df["time_value"], format='%Y%m%d') |
| 198 | + api_df.drop("direction", axis=1, inplace=True) |
| 199 | + api_df["data_source"] = data_source |
| 200 | + api_df["signal"] = signal |
| 201 | + |
| 202 | + return api_df |
0 commit comments