# Add Static Dataset

> Add a new static dataset to tilda-static-data geojson and integrate it through tilda-geo. Requires a matching tilda-static-data worktree, relinked geojson symlink, inferred config, missing-data questions, and type-guided meta.ts.

- Skill: `fixmyberlin/add-static-dataset` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add fixmyberlin/add-static-dataset`
- Raw SKILL.md: https://api.skillmd.com/api/skills/fixmyberlin/add-static-dataset/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: fixmyberlin (https://skillmd.com/u/fixmyberlin)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/fixmyberlin/add-static-dataset

---


# Add Static Dataset

Adds a new static dataset folder to **`tilda-static-data`** and uses the `tilda-geo` upload pipeline to process it. In `tilda-geo`, `app/scripts/StaticDatasets/geojson` is a symlink and must point at the matching `tilda-static-data` worktree before editing data files.

## When to Use

- User wants to add a new GeoJSON dataset
- User mentions adding static data, geojson files, "uploads", or datasets
- User provides a geojson file path and wants it integrated

## Required Information

Extract from user prompt or ask if missing:

1. **Group folder** (e.g., `region-berlin`, `region-bb`) - parent folder name
2. **Sub-folder name** (e.g., `berlin-bezirke`) - dataset folder name (must be valid slug). **Must include region/group prefix**: for `region-berlin` use `berlin-<name>`, not just `<name>`
3. **GeoJSON file path** - absolute path to source file on disk
4. **Regions** - array of region slugs (e.g., `['infravelo']`, `['berlin']`)
5. **Optional**: transformation needs, style preferences, category, attribution, license

## Process

### 0. Prepare `tilda-static-data` worktree and symlink

Static dataset files live in the sibling repo **`tilda-static-data`**, not in `tilda-geo`.

- If the task only reads existing static data, the regular symlink to `../../../../tilda-static-data/geojson` is enough.
- If the task adds or edits dataset files, create a matching `tilda-static-data` worktree and relink `geojson` there.

For a branch named `my-branch`:

```bash
cd ../tilda-static-data
git worktree add ../tilda-static-data--my-branch my-branch

cd ../tilda-geo--my-branch/app
rm -f scripts/StaticDatasets/geojson
ln -s ../../../../tilda-static-data--my-branch/geojson scripts/StaticDatasets/geojson
```

Do not edit `app/scripts/StaticDatasets/geojson` when it points at the regular shared `../tilda-static-data` checkout.

### 1. Create Folder Structure

```bash
app/scripts/StaticDatasets/geojson/<GROUP_FOLDER>/<SUB_FOLDER>/
```

**Important**: `/geojson` is a symlink. Before creating folders, ensure it points at the matching `tilda-static-data--my-branch` worktree.

Create directory if group folder doesn't exist. Ensure sub-folder name follows naming convention (see Required Information above).

### 2. Prepare GeoJSON file

Run these **in order** on committed data files. **Never** reproject, round, or pretty-print coordinates in `transform.ts` — use ogr2ogr (2.2–2.3) and oxfmt (2.4) on the files instead. Host GDAL 3.8+ (`brew install gdal`); not required for `bun run dev`.

#### 2.1 Move

Move source file to:

`app/scripts/StaticDatasets/geojson/<GROUP_FOLDER>/<SUB_FOLDER>/<FILENAME>.geojson`

This path should resolve into `../tilda-static-data--my-branch/geojson/...`.

#### 2.2 CRS → EPSG:4326

**Check:** Inspect the first coordinate pair after moving. For Germany, expect roughly lon **5–15**, lat **47–55**. Web Mercator (EPSG:3857) uses large meter values (e.g. ~1.5e6).

**Fix (only if not WGS84):** Reproject in place:

```bash
cd app/scripts/StaticDatasets/geojson/<GROUP_FOLDER>/<SUB_FOLDER>
ogr2ogr -f GeoJSON -t_srs EPSG:4326 -lco COORDINATE_PRECISION=8 \
  <FILENAME>.geojson.tmp <FILENAME>.geojson \
  && mv <FILENAME>.geojson.tmp <FILENAME>.geojson
```

#### 2.3 Coordinate precision (max 8 decimals)

**Check:** Coordinates should have at most **8** decimal places (matches upload pipeline and API exports).

**Fix (only if more than 8):** Round in place (file must already be EPSG:4326):

```bash
cd app/scripts/StaticDatasets/geojson/<GROUP_FOLDER>/<SUB_FOLDER>
ogr2ogr -f GeoJSON -lco COORDINATE_PRECISION=8 \
  <FILENAME>.geojson.tmp <FILENAME>.geojson \
  && mv <FILENAME>.geojson.tmp <FILENAME>.geojson
```

Upload runs `validateProjection` again on WGS84 bounds.

#### 2.4 Format (oxfmt)

From `app/` (skill / agents only — devs use format-on-save in the editor):

```bash
bun run format-static-datasets-geojson -- \
  scripts/StaticDatasets/geojson/<GROUP_FOLDER>/<SUB_FOLDER>/*.{json,geojson}
```

#### 2.5 Gzip (if large)

**Check:** Uncompressed `.geojson` size **after** formatting.

- **> 6 MiB** (6 × 1024² bytes): compress so the folder ships only the archive:

```bash
cd app/scripts/StaticDatasets/geojson/<GROUP_FOLDER>/<SUB_FOLDER>
gzip -9 -f <FILENAME>.geojson
```

`gzip` replaces the file with `<FILENAME>.geojson.gz` and removes the plain `.geojson` (`-9` max compression, `-f` overwrite).

### 3. Create transform.ts (if needed)

Only if transformation required (**not** for CRS or coordinate precision — Step 2 only). Use helpers from `app/scripts/StaticDatasets/geojson/_utils`:

- `transformUtils.ts` - property transformations
- `defaultLayerStyles.ts` - default styling helpers
- `translateUtils.ts` - translation helpers

Example:

```typescript
import { FeatureCollection } from 'geojson'

export const transform = (data: FeatureCollection) => {
  // Use helper functions from _utils when possible
  return data
}
```

### 4. Create meta.ts

**Critical**: Research similar datasets in same group folder first.

**Before writing meta.ts**:

1. List all datasets in `<GROUP_FOLDER>/` directory
2. Read 2-3 similar `meta.ts` files (similar geometry type, similar purpose)
3. Infer patterns:
   - Category from similar datasets
   - Attribution patterns
   - License conventions
   - Style patterns
   - Inspector settings
4. If unclear, ask user to confirm or provide missing data

**Required fields** (from `app/scripts/StaticDatasets/types.ts`):

- `regions`: RegionSlug[] (required)
- `public`: boolean (required)
- `dataSourceType`: 'local' (required)
- `attributionHtml`: string (required)
- `configs`: Array with at least one config (required)

**Optional on `export const data`**:

- `dataUpdatedNote`: string
- `dataSourceMarkdown`: string
- `licence`: License type (see types.ts)
- `licenceOsmCompatible`: 'licence' | 'waiver' | 'no'

**Config required fields**:

- `name`: string
- `inspector`: { enabled: boolean, ... } or { enabled: false }
- `layers`: Layer[] (required)

**Config optional fields**:

- `category`: string | null — grouping key; titles/order are managed in Admin → Statische Datensatz-Kategorien (DB).
- `description`: string
- `legends`: Legend[]

**Style/Legend**: If not specified, make best guess:

- Use `defaultLayerStyles()` from `_utils/defaultLayerStyles.ts` for simple cases
- For polygons: fill + outline
- For lines: colored line with width
- For points: circle markers
- Create appropriate legend entries

**Stripe / Schraffur (`fill-pattern`)**: For diagonal stripes over polygons, reuse the **existing map sprite** image id **`stripe_texture`** (same hatch as private Parkflächen in generated `parking_areas`; listed in `app/public/map-style/sprite.json`). Do not add a new sprite. Stack a **second** `fill` layer **above** the solid fill: `fill-color` `'rgba(0, 0, 0, 0)'`, `fill-opacity` around `0.5` (tune as needed), `fill-pattern` `'stripe_texture'`. Skip `fill-outline-color` on the stripe layer if the base fill already defines an outline. Use a `filter` on the stripe layer when only some features should be hatched; if exports mix boolean `true` and string `'true'`, either use an `any` / `match` filter in `meta.ts` or normalize values in `transform.ts` (step 3) so the filter stays simple.

**Example structure**:

```typescript
import { MetaData } from '../../../types'
import { defaultLayerStyles } from '../../_utils/defaultLayerStyles'

export const data: MetaData = {
  regions: ['infravelo'], // From user or infer from group folder
  public: true,
  dataSourceType: 'local',
  attributionHtml: 'Source Name', // Ask if unclear
  licence: 'DL-DE/ZERO-2.0', // Infer from similar datasets
  configs: [
    {
      name: 'Dataset Name',
      categoryKey: 'berlin/misc', // Infer from similar datasets
      inspector: { enabled: false }, // Default unless specified
      layers: defaultLayerStyles(), // Or custom layers
    },
  ],
}
```

### 5. Verify Command

Check `app/scripts/StaticDatasets/updateStaticDatasets.ts` for correct params:

- `--folder-filter=<SUB_FOLDER_NAME>` (matches full path, so sub-folder name works)
- `--env=dev` (or staging/production; omit to be prompted)

**One-click command** from `app/` (replace `<SUB_FOLDER_NAME>` with actual folder name; `static-datasets-update` uses the **repository root** `.env` via `bun --env-file=../.env`, same as `bun run dev`):

```bash
bun run static-datasets-update -- --folder-filter=<SUB_FOLDER_NAME> --env=dev
```

Note: `updateDownloadSources.ts` is for WFS downloads (requires downloadConfig.ts). Use `updateStaticDatasets.ts` for local GeoJSON files.

### 6. Verify TypeScript Compilation

**Always run this after creating or modifying TypeScript files** (meta.ts, transform.ts, or any imports). From `app/` (same as Step 5):

```bash
bun run type-check-deploy
```

This temporarily removes the `geojson` symlink, runs TypeScript type-checking, and restores the symlink. It simulates the Docker build environment where symlinks aren't available, ensuring your code will compile correctly during builds.

## Validation

Before completing:

1. ✅ Folder structure created
2. ✅ GeoJSON prepared (**Step 2.1–2.5**: moved, EPSG:4326, ≤8 decimals, oxfmt, gzip if >6 MiB)
3. ✅ `transform.ts` only if needed — **never** for CRS (Step 2); `meta.ts` / `transform.ts` via format-on-save or `format-static-datasets-geojson`
4. ✅ meta.ts follows type structure (check with TypeScript)
5. ✅ Similar datasets in group folder reviewed for patterns
6. ✅ Command verified and provided as one-click action
7. ✅ `bun run type-check-deploy` run successfully (see Step 6)

## References

- User-facing request template (DE, GitHub issue style): [ANFRAGE-STATISCHER-DATENSATZ.md](./ANFRAGE-STATISCHER-DATENSATZ.md)
- Types: `app/scripts/StaticDatasets/types.ts`
- Examples: `app/scripts/StaticDatasets/geojson/region-berlin/*/meta.ts`
- Utils: `app/scripts/StaticDatasets/geojson/_utils/`
- Docs: `docs/Features-Parameter-Deeplinks.md`
- Update script: `app/scripts/StaticDatasets/updateStaticDatasets.ts`