gudusoft.gsqlparser.pp2.island

Class MultiWordKeywordMerger

Object

gudusoft.gsqlparser.pp2.island.MultiWordKeywordMerger

public final class MultiWordKeywordMerger
extends Object

Identifies runs of consecutive tokens that form a multi-word SQL keyword (e.g. LEFT OUTER JOIN, UNION ALL, GROUP BY) so the lexical island pipeline can treat each run as a single keyword unit.

What it does and does NOT do

The merger is read-only: it returns a list of MultiWordKeywordMerger.Match spans and does not mutate any Pp2Token or its roles. (Role mutation is reserved for the designated annotator stages — S9 zone detector, S21 clause annotator, S33 overlay — not this slice.) Later pipeline stages consume the match list to know which token indices belong to one keyword.

Matching rules

• Allowlist, not word-count. Only phrases present in MultiWordKeywordTable merge. GROUP BY merges; LEFT JOIN does not (only the three-word LEFT OUTER JOIN form is in the table).
• Left-to-right, longest-match-at-each-position, non-overlapping. The scan walks tokens in source order; at each position the longest applicable phrase wins and its tokens are skipped. This is a local (per-position) greedy match, not a global longest-set optimisation — it mirrors the Delphi TSQLion lexical scan. The curated MultiWordKeywordTable allowlist contains no phrases where an earlier shorter match would preclude a later longer one, so the two agree here. If a future table introduces such overlaps, revisit this to a candidate-generate-then-select-by-length strategy.
• Directly consecutive tokens, case-insensitive text. A comment or any other token between two phrase words breaks the match (its text will not equal the expected keyword word, so it naturally fails). Quoted identifiers and string literals never match because their text carries the surrounding quotes.

Plan reference: §7.3/S18, §7.4/S18.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	MultiWordKeywordMerger.Match An immutable multi-word keyword span over a Pp2TokenStream: the inclusive token-index range and the canonical upper-cased phrase.

Constructor Summary

Constructors

Constructor and Description
MultiWordKeywordMerger()

Method Summary

Modifier and Type	Method and Description
List<MultiWordKeywordMerger.Match>	findMatches(Pp2TokenStream stream, EDbVendor vendor) Find all multi-word keyword spans in stream for vendor.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

MultiWordKeywordMerger

public MultiWordKeywordMerger()

Method Detail

findMatches

public List<MultiWordKeywordMerger.Match> findMatches(Pp2TokenStream stream,
                                                      EDbVendor vendor)

Find all multi-word keyword spans in stream for vendor.

Parameters:

stream - the token stream to scan; must not be null

vendor - the dialect whose phrase table to use; must not be null

Returns:

an immutable, source-ordered, non-overlapping list of matches (possibly empty); never null

Throws:

NullPointerException - if stream or vendor is null