Schema discovery

BlueTusk implements provider-neutral ADO.NET schema discovery for applications, libraries and diagnostic tools that inspect database metadata at runtime.

Connection collections

Call GetSchema() to list supported collections:

await using var connection = await dataSource.OpenConnectionAsync();
var collections = await connection.GetSchemaAsync();

V1 supports:

Collection	Requires open connection	Purpose
MetaDataCollections	No	Lists the available schema collections
DataSourceInformation	No	Provider and identifier behavior
DataTypes	No	Provider type metadata
Restrictions	No	Restriction positions for each collection
ReservedWords	No	PostgreSQL reserved words
Databases	Yes	Visible databases
Schemas	Yes	Visible schemas
Tables	Yes	Tables and views filtered by restrictions
Columns	Yes	Column metadata filtered by restrictions

Catalogue-backed collections require an open connection because visibility is defined by the authenticated PostgreSQL session.

Restrictions

Restrictions are positional and collection-specific. For example:

var tables = await connection.GetSchemaAsync(
    "Tables",
    [
        connection.Database,
        "public",
        "orders",
        "BASE TABLE",
    ]);

Use null for a restriction you do not want to constrain. Treat returned metadata as the current authenticated view of the catalogue, not a durable schema snapshot.

Reader column schema

After executing a command, use GetColumnSchema or GetSchemaTable:

await using var command = connection.CreateCommand();
command.CommandText = """
    SELECT id, created_at, payload
    FROM events
    WHERE tenant_id = @tenant_id
    """;
command.Parameters.Add(
    new BlueTuskParameter<Guid>("tenant_id", tenantId));

await using var reader = await command.ExecuteReaderAsync(cancellationToken);
var columns = await reader.GetColumnSchemaAsync(cancellationToken);
var table = await reader.GetSchemaTableAsync(cancellationToken);

Column metadata exposes the information available from PostgreSQL row descriptions and provider type resolution, including:

• column name and ordinal;
• CLR data type;
• provider type identity and name;
• nullability when known;
• size/precision/scale when known; and
• source relation/column metadata when PostgreSQL supplies it.

Unavailable metadata remains unavailable; BlueTusk does not invent key or origin information.

Intentional exclusions

CommandBehavior.SchemaOnly and CommandBehavior.KeyInfo are excluded in V1 and throw NotSupportedException. Silently treating either as CommandBehavior.Default would execute behavior the caller did not request.

Use a bounded query that returns no rows when you need PostgreSQL to describe a result shape:

SELECT id, created_at, payload
FROM events
WHERE false

This still parses, plans and describes an ordinary query, so apply the same SQL trust and timeout policy as any other command.

Tooling guidance

Schema discovery is useful for:

• Dapper-style mapping diagnostics;
• admin and inspection tools;
• code-generation inputs;
• compatibility probes; and
• generic data export.

It is not a replacement for EF Core reverse engineering, which preserves a larger PostgreSQL-specific model including indexes, constraints, policies, partitions, routines, extensions and other schema objects.

For database-first EF workflows, use the EF Core scaffolding guidance. For catalogue-driven provider types, read the type system guide.