fix(editor): match real F∆I YAML type shape + warn on bad types

The 0.15.0 type-token coloring regex hunted for `type: <name>`,
but the F∆I YAML actually shapes types as `<fieldname>: <name>`
under `inputs:` / `outputs:` blocks. Real-world flows
(`hello.yaml`, `extract.yaml`) and module manifests (echo,
text-summarize, …) never produce the `type:` keyword unless
the operator hand-authors the long form.

Fix the regex to match an indented `KEY: VALUE` line where the
value is one of `text|json|bytes|file|number|integer`. Leading
whitespace is required so top-level keys (`name: foo`,
`version: 0.1.0`) can't false-match. This handles both
shapes — the implicit `pdf: bytes` and the explicit
`type: bytes` (used by module schema v2 inputs lists) —
because either way the pattern boils down to "key colon known
type token".

Analyzer also grows a type-token check: any inputs/outputs
field whose value isn't a known type lights up as a warning
("Unknown input type 'byes' …"), modulo `$ref` expressions
(flow outputs) and empty values (operator is mid-keystroke).

Adds `test/flow_analyzer_test.dart` with seven cases covering
empty, valid hello, unknown capability, unknown type, parse
error, version-bare matching, and the empty-installed-list
silence path.

Bumps the package to 0.15.1.

Signed-off-by: flemming-it <sf@flemming.it>
This commit is contained in:
flemming-it 2026-06-04 02:24:29 +02:00
parent 885d2db4e1
commit 911f368362
4 changed files with 162 additions and 7 deletions

View file

@ -19,6 +19,8 @@
import 'package:flutter_code_editor/flutter_code_editor.dart';
import 'package:yaml/yaml.dart';
import 'wire_colors.dart';
class FlowAnalyzer extends AbstractAnalyzer {
/// Returns the names of capabilities the operator has installed.
/// A closure so the analyzer always sees the current list
@ -84,10 +86,44 @@ class FlowAnalyzer extends AbstractAnalyzer {
}
}
}
// Flag any inputs/outputs declaration whose value isn't a
// recognised type token. Flow YAML keeps each field as a
// `<name>: <type>` pair under `inputs:` / `outputs:` a
// typo like `pdf: byes` should light up rather than fail
// silently at run time.
_checkFieldTypes(doc['inputs'], 'input', issues);
_checkFieldTypes(doc['outputs'], 'output', issues);
}
return AnalysisResult(issues: issues);
}
void _checkFieldTypes(YamlNode? block, String kind, List<Issue> out) {
if (block is! YamlMap) return;
for (final entry in block.nodes.entries) {
final value = entry.value.value;
if (value is! String) continue;
// Skip reference expressions like `$inputs.name` or
// `$step.field` only flow `outputs:` use these and they
// aren't type declarations.
if (value.startsWith(r'$')) continue;
// Skip plainly empty values the operator is typing, not
// declaring a typo.
if (value.trim().isEmpty) continue;
if (!kKnownTypes.contains(value)) {
out.add(
Issue(
line: entry.value.span.start.line,
message:
'Unknown $kind type "$value". '
'Use one of: ${kKnownTypes.join(", ")}.',
type: IssueType.warning,
),
);
}
}
}
}
/// Strip the `@<version>` tail from a capability spec so

View file

@ -39,13 +39,25 @@ class FlowYamlCodeController extends CodeController {
analyzer = FlowAnalyzer(availableCapabilities: provider);
}
// Matches `type:` followed by (optional) quote + one of our
// known type names + (optional) closing quote. The captured
// group is the type name only we use the parent match span
// to locate it inside the document so we can recolour just
// the name (not the `type:` key itself).
// Matches an indented `key: value` line where the value is
// one of our known type tokens. This covers both shapes the
// FI YAML uses:
//
// inputs:
// pdf: bytes # field-name type
// count: number
//
// inputs:
// - name: pdf
// type: bytes # explicit `type:` row
//
// We require at least one leading whitespace so top-level
// keys (`name: foo`, `version: 0.1.0`) can't false-match.
// The captured group is the value we resolve its offset
// inside the document and recolour only that token.
static final RegExp _typeAssignment = RegExp(
r'''\btype\s*:\s*(?:['"])?([A-Za-z_][\w-]*)(?:['"])?''',
r'''^[ \t]+[A-Za-z_][\w-]*\s*:\s*(?:['"])?([A-Za-z_][\w-]*)(?:['"])?\s*(?:#.*)?$''',
multiLine: true,
);
@override