In notebooks

Open Graph from a notebook’s C# steps in OpenBI Studio: Graph, a graph made from the rows the model gives the step, with the everyday measures as tables; and Og, every one of Open Graph’s functions in C# terms.

Of the 1,162 entries of this reference, 513 functions are methods of Og in C# terms, 555 more (vectors, matrices, selectors, random numbers, handlers) are in Og.Native as C declares them, 68 types and 2 constants have their C# names, and 14 are for C alone (functions with a variable number of arguments, the cleanup stack’s macros), each with what to use instead. Every entry’s page says its C# name under In notebooks.

The library is the engine’s own Open Graph, the same snapshot, beside the engine (og.dll, libog.dylib on a Mac): a measure a step computes is the one NETWORK.* computes on a report page, number for number. A graph made from an OAX step holds only the rows the model gives that user: row-level security applies as it does on a page.

Notebooks run only in OpenBI Studio, on your own computer. OpenBI Server shows a published notebook read-only, as it was when it last ran, and never runs its code.

// the shops and the products they sell, from the model (the rows this user may see)
var rows = Dax("EVALUATE SELECTCOLUMNS(Sales, \"From\", Sales[Shop], \"To\", Sales[Product], \"Liters\", Sales[Liters])");
var g = Graph.FromTable(rows, "From", "To", weight: "Liters");

// the everyday measures, as tables with a Node column
Display(g.PageRank(), g.Communities("leiden"), g.CutPoints(), g.Similar("jaccard", top: 10));

// any other function of Open Graph: its C name without og_, in PascalCase
var (girth, circle) = Og.Girth(g);
var tree = Og.MinimumSpanningTree(g, g.Weights, Og.MstAlgorithm.Prim);
var (cuts, flows) = Og.GomoryHuTree(g, g.Weights);

Graph

Graph.FromTable builds a graph from a table of edges as the OAX NETWORK.* functions build theirs: the nodes numbered by value (numbers, then text in code-point order ignoring case, then dates, then FALSE and TRUE; order: NodeOrder.AsTheyCome numbers them as they first appear), text compared as OAX compares it, a pair’s rows added up into one edge (either way round when undirected), rows joining a node to itself, without an end or with a weight that is not above 0 left out, each edge weighed by its rows’ total (their count without a weight column); merge: false keeps every row as an edge. Graph.FromTwoSided builds a two-sided graph (customers and products) from two columns. Its methods give tables with a Node column (or From and To), seeded as NETWORK.* is: the same groups and the same drawing as on a report page. A graph shown under a step lists its edges; ToNetwork() draws it.

Og: every function

• Names. A function’s C name without og_, in PascalCase: og_pagerank is Og.Pagerank, og_community_leiden Og.CommunityLeiden. Its parameters keep their order and names. Enumerations and structs are inside Og too: Og.Neimode.Out, Og.ArpackOptions.
• Values. A graph (og_t) is a Graph; a vector an array (double[], long[], bool[]), a matrix a two-dimensional array, a list of vectors an array of arrays, a sparse matrix a SparseMatrix of triplets; a vertex or edge selector is long[] of ids, null for all; a file is its path.
• Results. What C writes into its output arguments comes back as the method’s value, a tuple when there are several (var (vector, value) = Og.Pagerank(g, ...)); an argument C reads and writes is a ref (a Leiden start, a layout’s start positions).
• Errors. A failure is an OpenGraphException with its Code (Og.Error), the C function and Open Graph’s own words: Open Graph’s og_neighbors: Vertex 9 is not in the graph. (OG_EINVVID). Warnings are written under the step.
• Callbacks. A handler (breadth-first search, cliques, motifs, isomorphisms) is a C# function returning true to go on and false to stop; an exception it throws comes out of the call, and interrupting the step stops the search.
• Threads and seeds. Og.Threads is the logical processors less one unless set; every answer is the same at any number. Og.Seed(n) seeds the random methods; Graph’s use NETWORK.*’s seeds (20261014; 20261018 for Sir, as NETWORK.SIR).
• As C has it. Og.Native.og_pagerank and the rest take pointers, in an unsafe block (graph.Pointer is its og_t); Og.Read reads a vector or a matrix. Use them only when you need Open Graph exactly as C has it: a mistake there can crash the notebook’s process instead of showing an error. OpenBI’s agent writes Graph and Og’s methods, and the C interface only when you ask for it. Memory a Graph holds is freed when it is disposed, or by the collector.
• Finding one. In a step, completion lists Og’s methods with these pages’ words; an agent asks analytics_catalog with api: "og_pagerank" (or "Og" for the list by area); Og.Find("og_pagerank") gives an entry’s C# name in a step.

Every entry in C#

By page of this reference: each entry’s C# name, or why only C has it.

Graph basics

Attributes

Deterministic generators

Random graphs

Components and structure

Paths and distances

Centrality

Similarity, transitivity and mixing

Community detection

Cliques, independent sets and colouring

Cycles, motifs and epidemics

Flows, cuts and traversal

Operators and bipartite graphs

Isomorphism

Layouts

Spatial graphs and numbers

Reading and writing files

Hierarchical random graphs and progress

Linear algebra and embeddings

Vectors

Matrices, stacks, queues and heaps

Sparse matrices

String vectors, lists and bit sets

Jacobian

Rollup

Threads

SIMD

Binary interface

Last updated 6 November 2026 · Open Graph 0.1.0 · OpenBI Studio 0.3.78