gudusoft.gsqlparser.pp2.render

Class RegionAssembler

Object

gudusoft.gsqlparser.pp2.render.RegionAssembler

public final class RegionAssembler
extends Object

Assembles per-region rendered texts into a single output string, bridging gaps between regions with the original inter-region trivia extracted from the SourceSpanLedger.

Why the ledger owns inter-region trivia

Every RegionRenderer intentionally omits the leading whitespace for the first token of its range (see RenderedRegion). The bytes between adjacent regions — whitespace, comments, blank lines, --BEGIN_NO_FORMAT/--END_NO_FORMAT blocks — live in the original source at positions [r_i.endOffset, r_{i+1}.startOffset). The assembler is the counterparty that places those bytes into the output, so they reach the caller byte-for-byte regardless of which renderer handled the surrounding statements.

Assembly model

Given an ordered, non-overlapping list of RenderedRegions:

1. Preamble — source[0, first.startOffset)
2. For each region r_i:
   1. Emit r_i.getText() (the renderer's output)
   2. Emit source[r_i.endOffset, r_{i+1}.startOffset) (inter-region gap; the final region skips this step)
3. Trailer — source[last.endOffset, source.length())

An empty regions list returns the entire source string unchanged — all bytes come from the preamble/trailer path.

Comment policy

Phase-2 MVP implements only CommentPolicy.PRESERVE: inter-region bytes are always emitted verbatim. CommentPolicy.REANCHOR and CommentPolicy.REFLOW log a notice and fall back to PRESERVE; the Phase-3 island pipeline will provide real semantics.

Defensive behaviour

Overlapping or out-of-order region offsets (which should never occur when StatementBoundaryDetector S11 produces the list) are detected and logged; overlap bytes are skipped rather than emitted twice. Out-of-bounds source-slice requests are also guarded — no byte is lost due to an offset arithmetic bug in the caller.

Plan reference: §7.3/S15, §7.4/S15, §10.4.

Constructor Summary

Constructors

Constructor and Description
RegionAssembler() Vendor-agnostic assembler (-- and /* comments only).
RegionAssembler(EDbVendor vendor) Assembler that recognises vendor's line-comment syntax when deciding whether a region begins with a (trailing) comment.

Method Summary

Modifier and Type	Method and Description
String	assemble(List<RenderedRegion> regions, SourceSpanLedger ledger, Pp2FormatOptions opts) Assemble rendered regions into a final output string.
String	assemble(List<RenderedRegion> regions, SourceSpanLedger ledger, Pp2FormatOptions opts, boolean breakStatements) Assemble rendered regions, with explicit control over statement-break normalization.

Methods inherited from class Object

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

Constructor Detail

RegionAssembler

public RegionAssembler()

Vendor-agnostic assembler (-- and /* comments only).

RegionAssembler

public RegionAssembler(EDbVendor vendor)

Assembler that recognises vendor's line-comment syntax when deciding whether a region begins with a (trailing) comment. In particular # is treated as a line comment for every dialect except the SQL Server family, where #name is a temp-table identifier rather than a comment.

Method Detail

assemble

public String assemble(List<RenderedRegion> regions,
                       SourceSpanLedger ledger,
                       Pp2FormatOptions opts)

Assemble rendered regions into a final output string.

Parameters:

regions - ordered (source position), non-overlapping list of rendered regions; empty list is allowed

ledger - the source-span ledger for the original SQL; SourceSpanLedger.getSource() is the byte authority

opts - pp2 format options; Pp2FormatOptions.commentPolicy governs how inter-region trivia is emitted

Returns:

the assembled output string; never null

Throws:

NullPointerException - if any argument is null

assemble

public String assemble(List<RenderedRegion> regions,
                       SourceSpanLedger ledger,
                       Pp2FormatOptions opts,
                       boolean breakStatements)

Assemble rendered regions, with explicit control over statement-break normalization. Identical to assemble(List, SourceSpanLedger, Pp2FormatOptions) except breakStatements overrides Pp2FormatOptions.breakStatementsOnNewLine. The engine uses this overload to re-assemble verbatim (breakStatements=false) as a content-preservation fallback when a break would alter the token stream.

Parameters:

breakStatements - when true, whitespace-only line-break-free gaps between statements become a single newline; when false, all inter-region trivia is emitted verbatim