These are not really competitors. Apache Calcite is a data-management framework — a query optimizer, relational algebra layer and federation toolkit behind Hive, Flink and Druid — and its parser exists to feed that pipeline. If you are building an engine, use Calcite; GSP has no optimizer and is not an alternative. General SQL Parser solves the other problem: reading SQL that already exists, faithfully, across 42 dialects, including stored procedures and dynamic SQL.
Calcite if you need to plan, optimise, rewrite or federate queries — it does things GSP simply cannot. GSP if you need to understand SQL that someone else wrote: vendor-faithful ASTs, procedural code, dynamic SQL, column-level lineage, and a redistribution contract if you ship it in a product you sell. Many teams run both, because they are different halves of the problem.
For a framework you assemble yourself, this matters more than usual: the integration is your code, and so is the liability. Apache-2.0 lets you ship Calcite. It also disclaims warranties.
| Apache Calcite (Apache-2.0) | GSP (commercial) | |
|---|---|---|
| Shipping it inside your commercial product | Permitted. Apache-2.0 lets you redistribute — with warranties expressly disclaimed and no indemnity to you or your customer. | A Distribution License: a named contract covering redistribution in your product, with Gudu carrying obligations rather than disclaiming them. |
| When a customer’s SQL breaks the parser | You file a JIRA ticket with the Apache project, or you fix it yourself in your fork. Both are real options; neither comes with a date. | You send the failing statement and it is scheduled into a release. |
| Who owns the integration burden | You do. Calcite is a framework: you assemble the parser, validator, catalog and planner into something that fits your product. | A product with an API. Less flexible by design, and less of your engineering time. |
| Procurement, security and legal review | No counterparty. Nobody to sign a DPA, complete a security questionnaire, or answer your customer’s auditor. | A company on the other side of a contract: security review, VPAT and audit questions get answered by a person. |
Calcite wins the first two rows outright. The categories barely overlap.
| Capability | Apache Calcite | GSP |
|---|---|---|
| What it is actually for | A data-management framework: relational algebra, cost-based query optimization and federation. It powers Hive, Flink and Druid. Its parser exists to feed that pipeline | A parsing and semantic-analysis SDK. It has no optimizer and no execution engine |
| Query planning, rewriting, federation | Yes — this is its reason to exist, and GSP cannot do any of it | No. If you need a planner, use Calcite and do not let this page distract you |
| License | Apache-2.0, free for commercial use | Commercial, with a 90-day free trial |
| Runs in-process on the JVM | Yes — free in-process parsing genuinely exists on the JVM | Yes, plus a native .NET edition |
| Dialect faithfulness | A configurable ANSI-leaning parser with dialect conformance settings — designed to normalise SQL toward relational algebra | 42 dialects, each with a dedicated grammar that preserves vendor-specific syntax rather than normalising it away |
| Stored procedures / procedural SQL | Out of scope — Calcite parses queries, not PL/SQL packages or T-SQL batches | Fully parsed ASTs: PL/SQL packages, types and triggers; T-SQL blocks, TRY/CATCH, EXECUTE; Db2 SQL PL cursors; Teradata procedures and BTEQ |
| Dynamic SQL | Not evaluated | Evaluated — EXEC(’…’) strings and parameter bindings folded to recover lineage |
| Column-level lineage | Derivable from RelNode metadata if you build it, and some projects do | A lineage engine across queries, procedures, dynamic SQL and multi-statement scripts, with impact analysis and call graphs |
| Offline validation without a catalog | Validation generally assumes a schema/catalog is available | Validation with error positions, no catalog or database connection required |
| SQL formatting | Via the target dialect writer | Dedicated formatter, 72+ style options |
Yes. If you need relational algebra, cost-based optimization, or federation across sources, Calcite is the right foundation and GSP is not an alternative to it — we have no optimizer and no execution layer. This page is only useful if what you actually need is to read and understand existing SQL rather than plan and execute it.
Yes, and we will not imply otherwise. Apache-2.0 permits redistribution. What it does not do is transfer risk: warranties are disclaimed, so when a customer hits a parse failure in production the cost and the deadline are yours. A Distribution License is about who carries that obligation rather than about permission. For internal platforms that is worth nothing; for software sold into regulated enterprises it is frequently the entire reason for the purchase.
Plenty of teams do, and it works. The reason to look elsewhere is a difference in intent rather than quality: Calcite normalises SQL toward relational algebra, which is exactly right for planning and slightly wrong for fidelity. If you need to know precisely what the original vendor syntax said — for migration, audit, or lineage you will defend to a regulator — a dialect-faithful parser fits better. If you are normalising anyway, that difference does not matter to you.
No. Calcite and JSqlParser both run in-process on the JVM for free, so we will not pretend otherwise. The in-process argument only applies on .NET, where the free options are T-SQL-only or syntax-only. On the JVM, judge us on dialect fidelity, procedural parsing, lineage and the support contract.
Yes, and it is a common shape: Calcite for planning and execution, GSP for reading the SQL that arrives from outside — legacy warehouse dialects, stored procedures, migration sources — and for lineage. They are solving different halves of the problem.