Graph thinking is rewriting the SQL playbook. Knowledge graphs and flexible schemas are nudging SQL toward graph-friendly workloads. One post shows how to construct knowledge graphs from NLP datasets [1].
From NLP-derived data to interconnected graphs, the idea is to turn language into graph structures that SQL can navigate. The discussion behind the NLP-to-graph approach demonstrates how semantic relationships can be organized for queries without abandoning SQL foundations [1].
A SQLite extension called Pandora argues that everything could be a database, showing how nearly any data can be packed into a database engine [2]. This riff on universal storage highlights the allure of letting traditional engines handle diverse formats without bespoke systems.
On the other side of the spectrum, LLKV bills itself as Arrow-Native SQL over Key-Value Storage, enabling SQL over backends that usually serve simple key-value lookups [3]. This approach supports graph-like queries on non-relational stores, reinforcing SQL as a substrate for graph workloads.
Together, these threads point to a future where SQL is a universal substrate for graph and schema-flexible workloads, from knowledge graphs built on NLP data to ultra-flexible database extensions and key-value-backed graph queries. The trend to blend graph realism with SQL pragmatism is the story to watch next.
References
Creating Knowledge Graphs from NLP Datasets
Discusses building knowledge graphs from NLP datasets, exploring graph technologies and graph databases.
View sourceA SQLite extension for the crazy ones because everything could be a database
A SQLite extension enables treating any object as a database, challenging usual boundaries and prompting debate on extensibility and integration.
View sourceLLKV: Arrow-Native SQL over Key-Value Storage
A crate enabling SQL queries over a key-value store using Apache Arrow, enabling relational queries on KV backends via bridge
View source