1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
use indexmap::IndexMap;
use regex::Regex;
use super::{
ExtractedSchema, Language, Result, SchemaExtractor, map_primitive_type, map_type,
parse_annotation_params,
};
use crate::schema::intermediate::{
IntermediateArgument, IntermediateField, IntermediateQuery, IntermediateType,
};
pub(super) struct PythonExtractor;
impl SchemaExtractor for PythonExtractor {
fn extract(&self, source: &str) -> Result<ExtractedSchema> {
let mut types = Vec::new();
let mut queries = Vec::new();
let type_re = Regex::new(r"@fraiseql\.type\(([^)]*)\)\s*\nclass\s+(\w+)")?;
let field_re = Regex::new(r"^\s+(\w+):\s*(.+?)\s*$")?;
let query_re = Regex::new(r"@fraiseql\.query\(([^)]*)\)\s*\ndef\s+(\w+)")?;
let lines: Vec<&str> = source.lines().collect();
// Extract types
for cap in type_re.captures_iter(source) {
let params = parse_annotation_params(&cap[1]);
let name = cap[2].to_string();
// Find class body: lines after "class Name:" that are indented
// Match ends after "class Name", skip to next line for body
let match_end =
cap.get(0).expect("regex group 0 is always Some on a successful match").end();
let body_start_line = source[..match_end].lines().count();
let mut fields = Vec::new();
for line in lines.iter().skip(body_start_line) {
// Skip blank lines and docstrings
let trimmed = line.trim();
if trimmed.is_empty()
|| trimmed.starts_with('#')
|| trimmed.starts_with("\"\"\"")
|| trimmed.starts_with("'''")
{
continue;
}
// Stop at next class/function/decorator at column 0
if !line.starts_with(' ') && !line.starts_with('\t') && !trimmed.is_empty() {
break;
}
if let Some(fcap) = field_re.captures(line) {
let field_name = fcap[1].to_string();
let type_str = fcap[2].to_string();
let (graphql_type, nullable) = map_type(Language::Python, &type_str);
fields.push(IntermediateField {
vector_config: None,
vector_distance: None,
name: field_name,
field_type: graphql_type,
nullable,
description: None,
directives: None,
// #1329: refusal-only on a field; `extract` never authors one.
function: None,
deprecated: None,
requires_scope: None,
on_deny: None,
authorize: None,
hierarchy: None,
unique: None,
localized: false,
});
}
}
let description = params.get("description").cloned();
types.push(IntermediateType {
subscription_policy: None,
sql_source: None,
name,
fields,
description,
implements: Vec::new(),
requires_role: None,
is_error: false,
is_input: false,
relay: false,
embedded: false,
subscribable_tables: None,
subscribable_pre_image: false,
// `extract` scrapes source text; no language has a relationship annotation for it
// to find, and it already omits `inject_params` and `subscribable_tables` on the
// same grounds. Authoring one goes through the SDK or `fraiseql.toml` (#1266).
relationships: Vec::new(),
inject_params: indexmap::IndexMap::new(),
});
}
// Extract queries
for cap in query_re.captures_iter(source) {
let params = parse_annotation_params(&cap[1]);
let name = cap[2].to_string();
let fn_end =
cap.get(0).expect("regex group 0 is always Some on a successful match").end();
// The real SDK derives the return type, list-ness and nullability
// from the function's return annotation (`-> list[Post]`,
// `-> Post | None`, `-> Post`) — it has no `return_type=` /
// `return_array=` decorator kwargs. The kwargs are kept as
// overrides so pre-SDK-dialect sources still extract.
let (ann_type, ann_list, ann_nullable) = parse_python_return_annotation(source, fn_end);
let return_type = params.get("return_type").cloned().or(ann_type).unwrap_or_default();
let returns_list =
params.get("return_array").map_or(ann_list, |v| v == "true" || v == "True");
let sql_source = params.get("sql_source").cloned();
// Parse function arguments (skip self, *, etc.)
let arguments = extract_python_query_args(source, fn_end);
queries.push(IntermediateQuery {
count: false,
name,
return_type,
returns_list,
nullable: ann_nullable,
arguments,
description: None,
sql_source,
sql_source_kind: fraiseql_core::schema::SourceKind::Relation,
// `extract` scrapes source text and produces no `functions` section, so it
// cannot produce the declaration a `function = "<name>"` binding must pair
// with (#1329). Authoring a function-backed field goes through the SDK.
function: None,
auto_params: None,
deprecated: None,
jsonb_column: None,
relay: false,
inject: IndexMap::default(),
exclude_inject_defaults: Vec::new(),
read_routing: fraiseql_core::db::types::ReadRouting::default(),
cache_ttl_seconds: None,
additional_views: vec![],
requires_role: None,
requires_actor: Vec::new(),
relay_cursor_type: None,
// `extract` scans SDK source text; it does not parse REST annotations.
rest: None,
rest_stream: false,
// `extract` scrapes source text; the pagination-order override (#1303) is an
// authoring decision, made through the SDK or `fraiseql.toml`. Absent, the
// compiler derives the entity identity, which is the right answer here.
pagination_order: None,
});
}
Ok(ExtractedSchema { types, queries })
}
}
/// Infer `(return_type, returns_list, nullable)` from the Python return
/// annotation following the function signature: `-> list[Post]` is a list,
/// `-> Post | None` is nullable, `-> Post` is a plain non-null result.
/// Returns `(None, false, false)` when no annotation is present.
fn parse_python_return_annotation(source: &str, fn_start: usize) -> (Option<String>, bool, bool) {
let rest = &source[fn_start..];
// Annotation lives between the signature's `->` and the `:` that opens the body.
let Some(arrow) = rest.find("->") else {
return (None, false, false);
};
let after_arrow = &rest[arrow + 2..];
let Some(colon) = after_arrow.find(':') else {
return (None, false, false);
};
let ann = after_arrow[..colon].trim();
if let Some(inner) = ann.strip_prefix("list[").and_then(|s| s.strip_suffix(']')) {
(Some(inner.trim().to_string()), true, false)
} else if let Some(t) = ann.strip_suffix("None").and_then(|s| s.trim_end().strip_suffix('|')) {
(Some(t.trim().to_string()), false, true)
} else if ann.is_empty() {
(None, false, false)
} else {
(Some(ann.to_string()), false, false)
}
}
pub(super) fn extract_python_query_args(
source: &str,
fn_start: usize,
) -> Vec<IntermediateArgument> {
let mut args = Vec::new();
// Find the function signature between parens
let rest = &source[fn_start..];
let Some(open) = rest.find('(') else {
return args;
};
let Some(close) = rest[open..].find(')') else {
return args;
};
let sig = &rest[open + 1..open + close];
let arg_re = Regex::new(r"(\w+):\s*(\w+)").expect("valid regex");
for cap in arg_re.captures_iter(sig) {
let name = &cap[1];
// Skip 'self' and bare '*'
if name == "self" {
continue;
}
let type_str = &cap[2];
let graphql_type = map_primitive_type(type_str);
args.push(IntermediateArgument {
name: name.to_string(),
arg_type: graphql_type,
nullable: false,
default: None,
description: None,
deprecated: None,
localized: false,
});
}
args
}