SQLise Blog

BigQuery and Snowflake feel similar until you port a query. Here are the dialect differences that cause the most breakage.

A few small macro habits keep your compiled SQL from turning into spaghetti — and keep lineage tools honest.

DISTINCT looks shorter. GROUP BY survives more edits. Here is the difference that matters in real codebases.

PIVOT and CASE WHEN do the same thing in different syntaxes. The right pick depends on dialect support and how dynamic your columns are.

If joins feel like guessing, this is the visual model that makes them click. Five join types, five tiny diagrams, no Venn diagrams.

EXPLAIN output is intimidating because it is dense, not because it is hard. Here is the order to read it in, and the three things actually worth fixing.

Recursive CTEs are the right tool for org charts, BOMs, and graph walks. They are also the easiest CTE to write infinite loops with.

SCD2 is the standard way to keep history of dimension attributes. Implementing it correctly takes a careful MERGE plus an INSERT.

An anti-join asks 'which rows in A are not in B?' Three idiomatic ways to write it; one common way that breaks on NULLs.

GROUP BY is the part of SQL most likely to silently lie to you. Here are the five classic ways it goes wrong and how to see them at a glance.

MERGE looks declarative but ships footguns. Here are the four bugs every team hits, and the patterns that prevent them.

NULL is not zero, not empty string, and not equal to itself. Here is the practical mental model that prevents the most common NULL bugs.

Snowflake has built-in object dependencies, but they don't tell you what feeds a specific column. Here are three ways to fix that.

CTEs read top-down, subqueries read inside-out. The right choice depends less on style and more on what you need the planner to do.

If your row counts mysteriously doubled overnight, the cause is almost always one of four join mistakes. Here is how to spot each one fast.

Window functions look intimidating because of the syntax, not the idea. Once you see the picture, OVER is just 'compute this across a moving window of rows'.