Skip to content
Back to skills

31 Parse Getinfo Results

ASecurity

Sub-skill of google-earth-engine: 3.1 Parse getInfo() Results (+2).

  • 17 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 9, 2026
datapythongo

Security analysis

A100/100

Scanned September 9, 2026

npx -y skills add vamseeachanta/workspace-hub --skill 31-parse-getinfo-results --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of 31 Parse Getinfo Results?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for 31 Parse Getinfo Results
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vamseeachanta-31-parse-getinfo-results/badge)](https://www.skillsdirectory.com/skills/vamseeachanta-31-parse-getinfo-results)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: google-earth-engine-31-parse-getinfo-results
description: 'Sub-skill of google-earth-engine: 3.1 Parse getInfo() Results (+2).'
version: 1.0.0
category: engineering/gis
type: reference
scripts_exempt: true
---

# 3.1 Parse getInfo() Results (+2)

## 3.1 Parse getInfo() Results


```python
# Always limit getInfo() to small results — expensive call
result = stats.getInfo()
depth_min = result["elevation_min"]
depth_max = result["elevation_max"]
depth_mean = result["elevation_mean"]
```


## 3.2 Read Exported GeoTIFF


```python
import rasterio
import numpy as np

with rasterio.open("gebco_north_sea_500m.tif") as src:
    depth = src.read(1).astype(float)
    depth[depth == src.nodata] = np.nan
    transform = src.transform
    crs = src.crs
print(f"Grid: {depth.shape}, depth range: {np.nanmin(depth):.0f} to {np.nanmax(depth):.0f} m")
```


## 3.3 Time-Series to DataFrame


```python
import pandas as pd

# Reduce collection to time series at a point
point = ee.Geometry.Point([-1.5, 57.0])
ts = era5_ws.map(lambda img: img.reduceRegion(
    reducer=ee.Reducer.mean(),
    geometry=point,
    scale=10000
).set("date", img.date().format("YYYY-MM-dd")))

data = ts.aggregate_array("date").getInfo()
ws_vals = ts.aggregate_array("wind_speed").getInfo()
df = pd.DataFrame({"date": data, "wind_speed_ms": ws_vals})
df["date"] = pd.to_datetime(df["date"])
```

---

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.