Rows without a match are dropped by semi join.

highlighted = computed this step

No-match rows are dropped

A left row with no matching right row has exists equal to no.

no match means drop\text{no match means drop}

Read dropped rows

2 left rows are dropped by the semi join.

dropped rows=2\text{dropped rows}=2

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

No-match rows are dropped: left tableemp_idnamedept_id1Ada102Ben203Cy204Dee305EliNULL No-match rows are dropped: right tabledept_iddept_name10ops20eng20platformNULLunknown candidate pairsleftSourcerightSourceleftKeyrightKeymatched00[10][10]yes01[10][20]no02[10][20]no03[10][NULL]no10[20][10]no11[20][20]yes12[20][20]yes13[20][NULL]no20[20][10]no21[20][20]yes22[20][20]yes23[20][NULL]no30[30][10]no31[30][20]no32[30][20]no33[30][NULL]no40[NULL][10]no41[NULL][20]no42[NULL][20]no43[NULL][NULL]no existence per left rowleftSourceleftKeymatchingRightSourcesexistskept0[10][0]yesyes1[20][1, 2]yesyes2[20][1, 2]yesyes3[30][]nono4[NULL][]nono semi join kept-left rowsemp_idnamedept_id1Ada102Ben203Cy20 semi join factsfactvalueleftRowCount5rightRowCount4candidatePairCount20matchingPairCount5leftRowsWithMatch3keptLeftRows3droppedLeftRows2duplicateRightMatches2nullKeyCandidatePairs8modesemirowOrderleft_source_order_once

Still no right output

Dropped rows disappear; right rows still are not output.

existence filter only\text{existence filter only}

Summary

Semi join is a left-row filter based on right-row existence.

left row filter\text{left row filter}