netload

A lightweight, dependency-light Python parser for .net network files (PTV Visum / Simetra and other .net text exports).

It reads the structured text tables of a .net file directly, without requiring PTV Visum to be installed, exposes every table as a Pandas DataFrame, and can transform any table (or the whole file) into CSV, JSON or GeoJSON.

Disclaimer: This project is independent and is not affiliated with, endorsed by, or supported by PTV Group or its partners. PTV Visum is a registered trademark of PTV Group.

---

Features

• Parses .net files generically: NODE, LINK, ZONE, TURN, USERATTDEF and any other table present in the file.
• No PTV Visum required.
• Every table is exposed as a real pandas.DataFrame.
• Automatic encoding detection (UTF-8 with/without BOM, Windows-1251/CP1251, UTF-16), so Russian and other non-ASCII content is read correctly.
• Robust to comments (*), blank lines, BOM markers and missing values.
• Export the whole file to CSV, JSON or GeoJSON in one call.
• Transform a single table to CSV / JSON / GeoJSON (via the API or the CLI).
• GeoJSON conversion uses WKT columns (WKTSURFACE, WKTPOLY, GEOMETRY) or XCOORD/YCOORD with a built-in WKT parser (no shapely needed).
• List every table with its fields (columns).
• Small command-line interface.

Installation

pip install netload

Quick example

from netload import read_net

network = read_net("model.net")

print(network.table_names)

print(network.nodes.head())

network.export_csv("./output")

Resulting files:

output/
├── NODE.csv
├── LINK.csv
├── ZONE.csv
└── TURN.csv

API

Reading

from netload import read_net

network = read_net("model.net")                 # auto-detect encoding
network = read_net("model.net", encoding="cp1251")
network = read_net("model.net", infer_types=True)  # light numeric inference

The returned Network object:

Member	Description
network.tables	dict of table name → Table (preserves file order)
network.table_names	list[str] of table names
network.has_table(x)	whether table x exists
network.get_table(x)	DataFrame of table x
network[x]	DataFrame of table x
network.nodes	DataFrame of NODE (if present)
network.links	DataFrame of LINK (if present)
network.zones	DataFrame of ZONE (if present)
network.turns	DataFrame of TURN (if present)
network.to_dict()	dict[str, pd.DataFrame]
network.export_csv()	writes one CSV per table (see below)
network.export_json()	writes one JSON per table (see below)
network.export_geojson()	writes GeoJSON for tables with geometry (see below)

Export CSV

network.export_csv("./output", sep=";", encoding="utf-8")

Tables without any data row are skipped by default (skip_empty=False to include them).

Export a single table (or a subset)

By default all tables are exported. Pass tables=[...] to export only specific tables:

network.export_csv("./output", tables=["NODE"])          # NODE.csv only
network.export_csv("./output", tables=["NODE", "LINK"])  # NODE.csv + LINK.csv
network.export_json("./output", tables=["ZONE"])
network.export_geojson("./output", tables=["ZONE", "LINK"])

The same applies to the CLI with --tables NODE,LINK.

Export JSON

network.export_json("./output")   # writes NODE.json, LINK.json, ZONE.json, ...

Each file contains a JSON array of records (Russian characters are kept as-is, not escaped).

Export GeoJSON

network.export_geojson("./output")  # writes NODE.geojson, ZONE.geojson, LINK.geojson, ...

Tables that have geometry are exported as GeoJSON FeatureCollection files:

• tables with a WKT column (WKTSURFACE, WKTPOLY, GEOMETRY) use its geometry (e.g. ZONE → MultiPolygon, LINK → LineString);
• tables with XCOORD/YCOORD become Point features (e.g. NODE);
• tables without any geometry are skipped.

Transform a single table

from netload.exporters import export_table_csv, export_table_json, export_table_geojson

export_table_json(network["NODE"], "node.json")
export_table_geojson(network["ZONE"], "zone.geojson")
export_table_csv(network["LINK"], "link.csv")

Errors

from netload import VisumNetError, VisumNetParseError, VisumNetEncodingError

• Missing file → FileNotFoundError
• Invalid structure → VisumNetParseError (includes file path and line number)
• Undecodable file → VisumNetEncodingError

CLI

netload model.net

Visum network loaded successfully

Tables:
- NODE: 1542 rows
- LINK: 3210 rows
- ZONE: 145 rows
- TURN: 872 rows

List tables with their fields:

netload model.net --list-tables

Tables and fields:
- NODE: 1542 rows
    fields: NO, NAME, CONTROLTYPE, T0PRT, ...

Preview a table (fields + first 5 rows by default):

netload model.net --preview NODE
netload model.net --preview NODE --rows 10

Export everything:

netload model.net --export-csv ./output
netload model.net --export-json ./output
netload model.net --export-geojson ./output

Export only specific tables (default: all):

netload model.net --export-csv ./output --tables NODE,LINK
netload model.net --export-json ./output --tables ZONE

Transform a single table:

netload model.net --table NODE --to-csv node.csv
netload model.net --table NODE --to-json node.json
netload model.net --table ZONE --to-geojson zone.geojson

Force a file encoding:

netload model.net --encoding cp1251 --list-tables

Tables

The parser is generic: it does not hard-code the set of supported tables. Any section starting with $ is parsed and made available, for example:

• NODE
• LINK
• ZONE
• TURN
• MAINTURN
• CONNECTOR
• LINKTYPE
• TSYS
• LINE, LINEROUTE, TIMEPROFILE, VEHJOURNEY, …
• USERATTDEF
• VERSION, NETWORK, …

If a file contains a table that is not listed here, it is parsed and exposed the same way.

How it works

1. The file bytes are read and the encoding is detected (BOM first, then UTF-8, then CP1251).
2. The text is scanned line by line.
3. A line starting with $NAME:COL1;COL2;... opens a new table with its columns.
4. Following non-comment, non-blank lines are parsed as ;-separated data rows.
5. Each table becomes a pandas.DataFrame (original string values are preserved by default).

Limitations (v0.1.0)

• Values are treated as ;-separated fields; quoted fields containing the separator character are not yet supported.
• Optional light numeric type inference is available via infer_types=True, but the default preserves the original string values to avoid any data corruption.
• Multi-line records and nested sub-tables are not supported yet.
• The built-in WKT parser covers POINT, LINESTRING, POLYGON, MULTIPOINT, MULTILINESTRING and MULTIPOLYGON (the geometry types used by .net files).

Development

python -m venv .venv
.venv/bin/pip install -e ".[dev]"
.venv/bin/python -m pytest

License

MIT. See LICENSE.

Disclaimer

This project is independent and not affiliated with PTV Group. PTV Visum is a registered trademark of PTV Group. This package is provided "as is", without any warranty.