'Monthly' object has no attribute '_types'

[clampr]

The following issue was reported on RapidAPI.

The user observed that passing a point's elevation/altitude causes some requests to fail with the error message:

'Monthly' object has no attribute '_types'

Apparently, this is a bug that occurs when no weather stations are found near the specified location.

We need to investigate if this issue only affects the Monthly class or perhaps also Hourly, Daily etc and provide a fix.

I will attach CoPilot's investigation report in a follow-up comment.

[clampr]

Investigation: alt Parameter Issues in Meteostat API v1

Investigation Request

Can you investigate if the described behavior can be reproduced in meteostat v1.7.4 directly or if the JSON API wrapper is causing the issue. Please comment in this issue with your findings (including the code used for querying + the response).

Findings Summary

After thorough investigation, I can confirm that the issues are caused by bugs in the underlying meteostat library (v1.7.4), not by the JSON API wrapper in this repository. The API wrapper correctly passes all parameters to the meteostat library.

Detailed Analysis

Issue 1: Requests fail without alt parameter but succeed with alt=0

Root Cause: When alt=None (parameter omitted), the meteostat Point class attempts to automatically compute the altitude from nearby weather stations:

# From meteostat/interface/point.py, lines 77-78
if self._alt is None:
    self._alt = stations.fetch().head(self.max_count)["elevation"].mean()

If no weather stations are found nearby:

• stations.fetch() returns an empty DataFrame
• Calling .mean() on an empty column returns nan
• The altitude becomes nan instead of staying None
• This can cause issues in subsequent altitude-based filtering

When alt=0 is explicitly provided, this automatic computation is skipped and the value remains 0.

Issue 2: Error "'Monthly' object has no attribute '_types'"

Root Cause: This is a bug in the meteostat library that occurs when no weather stations are found near the specified location.

In meteostat/interface/meteodata.py (line 121), the _get_data() method tries to create an empty DataFrame when no stations exist:

def _get_data(self) -> None:
    if len(self._stations) > 0:
        # Get list of datasets
        datasets = self._get_datasets()
        # Data Processing
        return processing_handler(datasets, self._load_data, self.processes, self.threads)

    # Empty DataFrame
    return pd.DataFrame(columns=[*self._types])  # ← BUG: _types doesn't exist!

The problem is that the _types attribute is only defined in the Stations class (meteostat/interface/stations.py, line 53):

class Stations(Base):
    _types: dict = {
        "id": "string",
        "name": "object",
        "country": "string",
        # ... etc
    }

But this attribute is not defined in:

• The MeteoData base class
• The TimeSeries class
• The Monthly, Daily, Hourly classes

Therefore, when these time series classes try to access self._types, they get an AttributeError.

Why Both Scenarios Can Fail

1. Without alt (None):

   • If there are no nearby stations within the 120km radius, alt becomes nan
   • When no stations pass the filters, len(self._stations) == 0
   • Calling data.normalize() triggers the _types error

2. With alt=0:

   • If there are no stations within the altitude range filter (default alt_range = 350m)
   • Again, len(self._stations) == 0
   • Same _types error occurs

Test Locations

The reported coordinates are in remote areas:

• Costa Rica: lat=9.4076, lon=-83.7135 - Rural area, may have sparse weather station coverage
• Kenya: lat=-1.2303795406138123, lon=36.91408397485202 - Near Nairobi, but altitude filtering may exclude stations

Code to Reproduce (Conceptual)

Since I don't have network access to test with live data, here's the code that would reproduce the issue:

from meteostat import Point, Monthly
from datetime import datetime

# Configure as in the API
Point.radius = 120000  # 120km radius
Monthly.threads = 4
Monthly.autoclean = False
Monthly.max_age = 60 * 60 * 24 * 7

# Test Case 1: Location with sparse station coverage, no alt
location = Point(9.4076, -83.7135, None)

# This will fail with AttributeError when no stations are found
try:
    data = Monthly(location, datetime(2024, 1, 1), datetime(2024, 12, 31), model=True)
    if data.count() > 0:
        data = data.normalize()  # ← Error occurs here if count == 0
        df = data.fetch()
except AttributeError as e:
    print(f"Error: {e}")  # 'Monthly' object has no attribute '_types'

# Test Case 2: Same location with alt=0
location2 = Point(9.4076, -83.7135, 0)

# May also fail depending on altitude filtering
try:
    data2 = Monthly(location2, datetime(2024, 1, 1), datetime(2024, 12, 31), model=True)
    if data2.count() > 0:
        data2 = data2.normalize()
        df2 = data2.fetch()
except AttributeError as e:
    print(f"Error: {e}")  # Same error if no stations found

API Wrapper Code Review

The API wrapper in this repository (app/data/point/monthly.py) correctly passes all parameters:

def point_monthly(
    lat: float,
    lon: float,
    start: datetime,
    end: datetime,
    alt: Optional[int] = None,  # ← Correctly optional
    model: bool = True,
    freq: Optional[str] = None,
    units: Optional[str] = None,
):
    # ...
    location = Point(lat, lon, alt)  # ← Correctly passes alt (None or value)
    data = Monthly(location, start, end, model=model)
    # ...

The wrapper properly:

• Declares alt as Optional[int] = None
• Passes it through to Point() unchanged
• Has proper error handling with try/except

Conclusion

The JSON API wrapper is working correctly. The issues are bugs in the meteostat library v1.7.4:

1. Bug Evaluate SQLite Caching #1: Missing _types attribute in time series classes causes AttributeError when no stations are found
2. Bug Consider splitting hourly data dumps into annual chunks #2: Setting alt=nan when no stations exist (instead of keeping None or using 0)

Recommendations

1. Report this bug to the meteostat library maintainers with the following fix suggestions:

   • Define _types in the base MeteoData class using _processed_columns
   • OR handle the empty DataFrame case differently when no stations are found
   • Better handling of alt=None when no stations exist

2. Optional workaround in the API wrapper: Add more defensive error handling:

   if data.count() == 0:
       raise HTTPException(
           status_code=404,
           detail="No weather stations found near the specified location."
       )

   However, this won't prevent the AttributeError during normalize() if called on empty data.

3. Update documentation: Clarify that some remote locations may not have weather station coverage, and that the alt parameter behavior depends on station availability.

Files Examined

Repository files:

• app/data/point/monthly.py
• app/data/point/daily.py
• app/data/point/hourly.py
• app/data/point/normals.py
• app/main.py

Meteostat library files (v1.7.4):

• meteostat/interface/point.py (lines 50-100)
• meteostat/interface/meteodata.py (lines 106-122)
• meteostat/interface/monthly.py
• meteostat/interface/timeseries.py
• meteostat/interface/stations.py (line 53)
• meteostat/series/normalize.py

[clampr]

Please fix linter issues (see PR comment) @copilot.