Choose data for a graph

Advanced: assumes basic Python, tabular data, and graph queries. For a guided introduction, start with Your first graph.

Start here after Your first graph when you want data for your own question. GraphForge v0.6.0 does not ship a graphforge.datasets catalog, GraphForge.from_dataset(), or dataset-download CLI commands. Download and inspect source files separately, then construct the graph through the public APIs.

Choose data that can answer the question

Before loading, write down what one record represents, what a relationship means, which people or documents are included, and what is missing. An edge list usually contains only connections; people with no recorded connections may be absent. A missing edge is not automatically evidence that no relationship exists.

Keep the source URL, citation, download date, file version or checksum, license, and any selection or cleaning steps. Counts from an example network describe that captured network, not a representative sample of a wider population. Interview excerpts and coding decisions need their own source links; a network file does not supply that qualitative context.

External sources

Source	What to inspect
SNAP	Network definitions, collection period, direction, and the dataset’s own citation
NetworkRepository	File format, node and edge meanings, and the original study
Neo4j examples	Teaching datasets and queries; check syntax and dependencies before reuse
LDBC benchmarks	Specialist engine-evaluation workloads, not a first research-project dataset

For a small runnable real-network example, the repository’s visualization examples fetch a specific Karate Club network and build it through public APIs. They require a source checkout and additional packages.

Bring your own files

For a small spreadsheet, export CSV and read it with Python’s csv.DictReader. Decide which columns identify records, which become properties, and which link records. Preserve identifiers as strings when leading zeros matter. Handle blank cells and repeated rows explicitly before creating nodes.

Use add_node and add_edge for small inputs as shown in Graph construction. Keep a mapping from your source identifiers to returned node handles so relationships connect the intended records. Bulk APIs are optional for larger imports; they do not decide what the data means or whether your sample is complete.

Check loaded counts and a few source-linked rows before analysis. Keep original files and the construction script. For persistence, follow save and reopen; for a bounded conclusion and later reuse, continue with Record an inquiry.