core: one type row per expression, canonicalized per dialect, across every engine - #4618

Merged

kyleconroy merged 16 commits into

mainfrom

claude/kind-dijkstra-88hp13

Sep 11, 2026

Commits on Sep 10, 2026

core: evaluate how types are represented, and design full type expressions in the catalog
claude
committed
e471171

Commits on Sep 11, 2026

core: review each engine's own type catalog, and canonicalize types on the way in
claude
committed
96c085e
core: PostgreSQL's canonical type names are format_type's
claude
committed
9dcdbf7
core: add a worked example of an array of arrays to the types note
claude
committed
b4271e2
core: one type row per expression, canonicalized, with the analyzer carrying the expression
claude
committed
895ed01
postgresql: typmods, dimensions, interval fields, declared types and namespaces in the core
claude
committed
fedc4a1
mysql: unsigned families, typmods, enum and set members, and the types MySQL reports for casts
claude
committed
48a9f7a
sqlite: a declared spelling stands on its affinity
claude
committed
14c2a58
clickhouse: casts, typed placeholders, aggregate-function idents, canonical forms and Nested columns
claude
committed
efc83d3
duckdb, googlesql, mssql: nested types, parameters, MAX, defaults and alias types in the core
claude
committed
4144f5d
docs: describe the types analyze reports and how the design landed
claude
committed
b0e1cf0
core: a type in another namespace resolves there and is reported qualified
claude
committed
85b92aa
core: express what a dialect does to a type as data rather than registered code
claude
committed
271b10f
core: answer the review: canonical type field, label parsing, namespaces, lookups
claude
committed
c70c5d6
core: answer the second review: qualified keys, nested lookups, rewrites after aliases
claude
committed
33ee062
endtoend: regenerate the process plugin golden for the PostgreSQL relation seed
claude
committed
fc7b02c

Navigation Menu

Uh oh!

core: one type row per expression, canonicalized per dialect, across every engine - #4618

core: one type row per expression, canonicalized per dialect, across every engine#4618
kyleconroy merged 16 commits into
mainfrom
claude/kind-dijkstra-88hp13

Commits on Sep 10, 2026

core: evaluate how types are represented, and design full type expressions in the catalog

Commits on Sep 11, 2026

core: review each engine's own type catalog, and canonicalize types on the way in

core: PostgreSQL's canonical type names are format_type's

core: add a worked example of an array of arrays to the types note

core: one type row per expression, canonicalized, with the analyzer carrying the expression

postgresql: typmods, dimensions, interval fields, declared types and namespaces in the core

mysql: unsigned families, typmods, enum and set members, and the types MySQL reports for casts

sqlite: a declared spelling stands on its affinity

clickhouse: casts, typed placeholders, aggregate-function idents, canonical forms and Nested columns

duckdb, googlesql, mssql: nested types, parameters, MAX, defaults and alias types in the core

docs: describe the types analyze reports and how the design landed

core: a type in another namespace resolves there and is reported qualified

core: express what a dialect does to a type as data rather than registered code

core: answer the review: canonical type field, label parsing, namespaces, lookups

core: answer the second review: qualified keys, nested lookups, rewrites after aliases

endtoend: regenerate the process plugin golden for the PostgreSQL relation seed