Parsing file formats, databases, data mining, and enrichment.
How to convert a website to JSON in 2026: read the JSON-LD or hidden API the page already ships, fall back to selectors, with a measured parser benchmark and hosted API prices checked 10 August 2026.
Back up an MS SQL Server database properly: recovery models, the T-SQL BACKUP command with ZSTD compression in SQL Server 2025, backup to S3 and Azure Blob, Agent and Ola Hallengren scheduling, and a restore drill. Versions and prices read 10 August 2026.
Big data and data mining checked against primary sources in August 2026: what each term means, where the single-machine line actually sits now, which Hadoop-era Apache projects have been retired, and current BigQuery, Snowflake, Redshift and S3 prices.
Apache Cassandra explained from the primary sources: masterless architecture, query-first data model, the operational costs nobody puts on a feature page, and what versions, prices and users looked like on 10 August 2026.
Clustering explained with measurements rather than adjectives: the five families, k-means and HDBSCAN in scikit-learn 1.8, why the elbow method should be retired, the O(n²) wall in hierarchical clustering and silhouette scoring, and what runs at scale in August 2026.
Data mining discovers patterns in data you already have; machine learning predicts on data it has not seen. Six terms pinned down, a comparison table, two traps you can reproduce in a minute, and tool release dates read from the projects themselves in August 2026.
What data enrichment is, how the matching key decides your match rate, what registries and enrichment APIs really cost with every price read from the vendor on 10 August 2026, what breaks at 40,000 rows, and where the law stands after the California deletion platform went live.
Data, information and knowledge are three states of the same material, and each step between them is separate work. Worked examples, the real history of the DIKW pyramid including the tier almost every retelling drops, the 2009 critique that undermines it, and what the ladder costs in bandwidth at ten thousand pages a night. Sources read on 10 August 2026.
OpenRefine for cleaning scraped data, rechecked in August 2026 against release 3.10.1: how the fingerprint clusterer really works, what GREL match() actually returns, and the three ways your data leaves a local-only tool.