Blog

Column cardinality is one of the biggest factors in Power BI model size and query speed. This post explains why, and covers practical techniques for reducing it.

Migrating from V4 to V5 is straightforward with Roslyn analyzers and code fixes. V4 isn't going away - both engines are maintained. Plus: 10 production analyzers to help you write correct, high-performance code.

Corvus.Text.Json.Toon provides bidirectional TOON conversion - a compact text format that removes repeated property names and punctuation from JSON, reducing token count for LLM prompts with zero-allocation UTF-8 APIs.

V5 extends the JSON type system with UTF-8 URI and IRI parsing, arbitrary-precision numerics via BigNumber, and first-class NodaTime date/time integration. All operating directly on the raw UTF-8 bytes.

Corvus.Text.Json V5 includes Utf8JsonPointer, a readonly ref struct that resolves RFC 6901 JSON Pointer paths directly against the pooled-memory document model with no allocation.

Corvus.Text.Json.Patch implements RFC 6902 JSON Patch with a fluent PatchBuilder and all six operations. Patches operate directly on the mutable document model with no intermediate copies.

Corvus.Text.Json.Yaml converts YAML 1.2 to JSON with 100% yaml-test-suite conformance using a custom ref struct tokenizer - no intermediate object model, no allocations on the hot path.

Corvus.Text.Json.JsonLogic provides a safe, side-effect-free rule engine that evaluates JSON-encoded business logic on average 3× faster than JsonEverything - with zero allocations.

Corvus.Text.Json.JMESPath implements the full JMESPath spec with 100% conformance, zero-allocation hot paths, and on average 28× faster than JmesPath.Net across 21 benchmarks.

Corvus.Text.Json.Jsonata brings the full JSONata language to .NET - 100% test suite conformance, on average 2× faster than Jsonata.Net.Native with 90–100% less memory, plus a source generator for compile-time code generation.

The standalone evaluator generates a lightweight validator with fully compliant JSON Schema annotation collection - ideal for form generators, schema-driven UIs, and documentation tools.

JsonDocumentBuilder and JsonWorkspace provide pooled, version-tracked mutable documents - the core V5 design trade-off that replaces V4's immutable-functional model with in-place mutation.

ParsedJsonDocument<T> uses ArrayPool-backed memory for just 136 bytes of GC pressure per document - 91% less than JsonNode - while providing a familiar System.Text.Json-compatible API.

Corvus.Text.Json V5 validates JSON against all major schema drafts over 10× faster than other .NET validators - with a simple EvaluateSchema() API and optional detailed diagnostics.

Annotate a partial struct with [JsonSchemaTypeGenerator] and get strongly-typed properties, validation, serialization, and mutable builders - all generated at build time with full IntelliSense.

Corvus.Text.Json V5 is a new engine for high-performance JSON in .NET - pooled-memory parsing, mutable documents, source-generated types, and three query languages. Here's why we built it.

VertiPaq fits millions of rows in memory by compressing columns. Learn how value, hash and run-length encoding work, and what they mean for model performance.

VertiPaq stores Power BI data column-by-column rather than row-by-row. That makes aggregations fast and cross-column queries the trickiest part of DAX.

Optimising DAX in Power BI: VertiPaq storage, compression, model design, how queries run through the storage, formula engines, and performance patterns.

Variable libraries in Microsoft Fabric manage environment-specific configuration. Learn where you can use them and how to set safer defaults for Dev/Test/Prod.

AI is like introducing a powerful new instrument to an orchestra. It creates possibilities that didn't exist before. But it still requires musicians who can read music, a conductor with a vision, rehearsal time, and the discipline to play together. An orchestra that lacks these fundamentals won't be saved by a new instrument — they'll just make new kinds of noise.

SharePoint isn't going away. This post compares five ways to ingest SharePoint data into Microsoft Fabric, with guidance on cost, complexity, and incremental loading.

Top-down AI strategy and bottom-up experimentation both fail alone: leading organisations combine them to drive real business results.

SQLBits is one of the largest data platform conferences in Europe. Here's a recap of my experience at SQLBits 2026, held at the ICC Wales.

Databricks SQL cold starts kill web API performance. Fix it with two-layer caching: Azure Blob Storage & IMemoryCache, using the Decorator pattern.