Error Recovery

What you'll learn

• What the parser produces for malformed source and how to inspect it.

Complete example

examples/error_recovery.zig:

for ([_][]const u8{ "1 + * 2", "(1 + 2", "1 @ 2" }) |source| {
    var tree = try parser.parseString(source);
    defer tree.deinit();
    std.debug.print("{s} => has_error={} nodes={d}\n", .{ source, tree.hasError(), tree.nodeCount() });
}

Expected output

1 + * 2 => has_error=true nodes=11
(1 + 2 => has_error=true nodes=15
1 @ 2 => has_error=true nodes=6

How it works

• Unexpected tokens are skipped and the skipped span becomes an ERROR leaf (named, isError(), hasError() set). Skipping accrues an error cost per byte, bounded by a maximum recovery budget.
• Unexpected end of input inserts a zero-width MISSING token for the expected symbol (e.g. the missing ) in (1 + 2) and continues parsing, so unclosed constructs still produce structured trees.
• Unrecoverable states fall back to wrapping all fragments in a root node marked hasError(), so a tree is always returned — parsing never fails just because the source is broken.
• tree.hasError() / node.hasError() flag tainted subtrees; walk children for isError() / isMissing() to pinpoint problems for diagnostics.

API used

• Tree — hasError; Node — isError, isMissing, hasError; Errors.

Related guides

• Testing for asserting recovery behavior, Troubleshooting.
• Error Recovery, Upstream Conformance.