Skip to content

Instantly share code, notes, and snippets.

View RaczeQ's full-sized avatar
🌍
Crafting new Open Source tools 🔮

Kamil Raczycki RaczeQ

🌍
Crafting new Open Source tools 🔮
View GitHub Profile
@RaczeQ
RaczeQ / duckdb_h3_only_query.sql
Created May 11, 2026 09:26
Compacted H3 cells to ranges of DuckDB query
SELECT
*
FROM
read_parquet ('s3://.../**/*.parquet')
WHERE
country = 'PL' -- hive partition filter
AND h3_res_3 IN (590501584509599743) -- hive partition filter
AND ( -- statistics filter pushdown from compacted cells
h3_res_12 BETWEEN 631033929750544895 AND 631033929750547967 -- H3 parent cell 626530330123177983
OR h3_res_12 BETWEEN 631033929750548991 AND 631033929750552063 -- H3 parent cell 626530330123182079
@RaczeQ
RaczeQ / h3_to_ranges.py
Last active May 11, 2026 09:32
Compacted H3 cells to ranges of expected resolution.
def compacted_h3_cells_to_ranges(
compacted: list[int], h3_resolution: int = 12
) -> list[tuple[int, int]]:
"""
Compacted H3 cells -> list of (r_min, r_max) integer ranges.
Each pair tightly brackets all descendants of the given cell at the same resolution.
Ranges across cells are disjoint (H3 bit layout property).
"""
RES_SHIFT = 52
import math
from itertools import pairwise
import geopandas as gpd
import numpy as np
from shapely import LineString, Point, Polygon, distance
from shapely.coords import CoordinateSequence
def locate_farthest_intersection_point(
import numpy as np
import geopandas as gpd
# keep a geodataframe of all points creating the edges
all_points = gpd.GeoSeries(
clipped_edges_gdf.get_coordinates(ignore_index=True).apply(
lambda row: Point(row["x"], row["y"]), axis=1
),
crs=4326,
)
# this is a code snippet, whole logic with checks is in the final notebook
subgraph_edges = ox.graph_to_gdfs(subgraph, nodes=False, edges=True)
# find all endpoints and check their edges outside. Clip edges exactly at the distance point.
edges_to_clip = {}
for node in set(subgraph.nodes).union([center_node]):
# iterate all edges starting from this node
for u, v, data in G.edges(node, keys=False, data=True):
# if whole edge is inside clipped subgraph - skip it
if v in subgraph:
@RaczeQ
RaczeQ / Snowflake_OvertureMaps_Pubs.sql
Last active January 19, 2026 10:29
Which three places in the UK have the highest density of pubs?
-- Select all pubs in the Great Britain and calculate how many pubs are in a 5 km radius around each pub.
WITH pubs AS (
-- irish_pub....................................eat_and_drink > bar > irish_pub
-- pub..........................................eat_and_drink > bar > pub
SELECT
id,
JSON_EXTRACT_PATH_TEXT(names, 'primary') as name,
JSON_EXTRACT_PATH_TEXT(categories, 'primary') as category,
confidence,
@RaczeQ
RaczeQ / Snowflake_OvertureMaps_Buildings.sql
Created January 13, 2026 07:56
Count number of buildings near green areas (600 meters) for Berlin, London and Paris
WITH division_ids AS (
SELECT
id,
CASE
WHEN country = 'DE' THEN 'Berlin'
WHEN country = 'FR' THEN 'Paris'
ELSE 'London'
END as city_name
FROM
OVERTURE_MAPS__DIVISIONS.CARTO.DIVISION
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8" />
<title>Trip animation – sequences (deck.gl + MapLibre)</title>
<meta name="viewport" content="width=device-width,initial-scale=1" />
<link href="https://unpkg.com/maplibre-gl@5.12.0/dist/maplibre-gl.css" rel="stylesheet" />
<!-- <link rel="stylesheet" href="https://fonts.googleapis.com/icon?family=Material+Icons" /> -->
<script src="https://unpkg.com/maplibre-gl@5.12.0/dist/maplibre-gl.js"></script>
@RaczeQ
RaczeQ / pyarrow_multiprocessing_streaming.py
Last active May 19, 2024 19:41
Pyarrow Multiprocessing with streaming the result
import multiprocessing
from pathlib import Path
from queue import Queue
from time import sleep
from typing import Callable
import pyarrow as pa
import pyarrow.parquet as pq
from tqdm import tqdm