FULL OUTER JOIN starts from two input tables.

highlighted = computed this step

Left and right tables

A FULL OUTER JOIN starts with two visible input tables.

two inputs\text{two inputs}

Read inputs

The left table has 4 rows; the right table has 5 rows.

left rows=4\text{left rows}=4

Full-outer-join examples are tiny finite-table transforms; SQL dialect completeness, type coercion, optimizer behavior, execution cost, indexing, output ordering guarantees, and database-product claims are out of scope.

Left and right tables: left tableemp_idnamedept_id1Ada102Ben203Cy304DeeNULL Left and right tables: right tabledept_iddept_name10ops20eng20platform40legalNULLunknown candidate key matchesleftSourcerightSourceleftKeyrightKeymatched001010yes011020no021020no031040no0410NULLno102010no112020yes122020yes132040no1420NULLno203010no213020no223020no233040no2430NULLno30NULL10no31NULL20no32NULL20no33NULL40no34NULLNULLno FULL OUTER JOIN output rowsleft.emp_idleft.nameleft.dept_idright.dept_idright.dept_name1Ada1010ops2Ben2020eng2Ben2020platform3Cy30NULLNULL4DeeNULLNULLNULLNULLNULLNULL40legalNULLNULLNULLNULLunknown output row classificationclassleftSourcerightSourcerowmatched00[1, Ada, 10, 10, ops]matched11[2, Ben, 20, 20, eng]matched12[2, Ben, 20, 20, platform]left_only2NULL[3, Cy, 30, NULL, NULL]left_only3NULL[4, Dee, NULL, NULL, NULL]right_onlyNULL3[NULL, NULL, NULL, 40, legal]right_onlyNULL4[NULL, NULL, NULL, NULL, unknown] FULL OUTER JOIN factsfactvalueleftRowCount4rightRowCount5candidatePairCount20matchingPairCount3leftOnlyRowCount2rightOnlyRowCount2outputRowCount7duplicateMatchRows1nullKeyCandidatePairs8rowOrdermatched_left_source_then_left_only_then_right_only

Both sides can survive

Matched rows, left-only rows, and right-only rows can all appear in the output.

both sides preserved\text{both sides preserved}

Summary

The rest of the book shows how each output class is built.

classes next\text{classes next}