feat(native): run metadata checks natively behind the preview engine

crates/sqlbuild-analysis/src/semantic_checks/_helpers/columns.rs → crates/sqlbuild-analysis/src/semantic_checks/_helpers/explanation/columns.rs

@@ -2,7 +2,7 @@
 
 use std::cmp::Ordering;
 
-use crate::semantic_checks::_helpers::text::{ascii, casefold};
+use crate::semantic_checks::_helpers::sql_text::text::{ascii, casefold};
 use crate::semantic_checks::constants::{MAX_EDIT_DISTANCE, MIN_ABBREVIATION_LENGTH};
 use crate::semantic_checks::models::SemanticDeferral;
 

crates/sqlbuild-analysis/src/semantic_checks/_helpers/completion.rs → crates/sqlbuild-analysis/src/semantic_checks/_helpers/explanation/completion.rs

@@ -5,19 +5,21 @@ use std::sync::LazyLock;
 
 use regex::Regex;
 
-use crate::semantic_checks::_helpers::bindings::{BindingOwner, binding_positions};
-use crate::semantic_checks::_helpers::columns::closest_column;
-use crate::semantic_checks::_helpers::explain::{
+use crate::semantic_checks::_helpers::explanation::columns::closest_column;
+use crate::semantic_checks::_helpers::explanation::explain::{
     Binary, ExplainContext, ModelEvidence, Shapes, binary_index, explain_model, is_explained_code,
     is_type_code,
 };
-use crate::semantic_checks::_helpers::messages::{
+use crate::semantic_checks::_helpers::explanation::messages::{
     compiled, missing_column, pattern, semantic_help, sentence_message,
 };
-use crate::semantic_checks::_helpers::opt_outs::opt_out_diagnostic;
-use crate::semantic_checks::_helpers::parsed_sql::{ParsedModelFacts, parsed_model_facts};
-use crate::semantic_checks::_helpers::poison::Poison;
-use crate::semantic_checks::_helpers::text::{LineIndex, ascii, plain_text};
+use crate::semantic_checks::_helpers::explanation::opt_outs::opt_out_diagnostic;
+use crate::semantic_checks::_helpers::recovery::bindings::{BindingOwner, binding_positions};
+use crate::semantic_checks::_helpers::recovery::poison::Poison;
+use crate::semantic_checks::_helpers::sql_text::parsed_sql::{
+    ParsedModelFacts, parsed_model_facts,
+};
+use crate::semantic_checks::_helpers::sql_text::text::{LineIndex, ascii, plain_text};
 use crate::semantic_checks::constants::{
     MODEL_RESOURCE_TYPE, SEMANTIC_CODE_PREFIX, SQL_TEST_COLUMN_CODE, SQL_TEST_COLUMN_PATTERN,
     UNKNOWN_COLUMN_CODE,

crates/sqlbuild-analysis/src/semantic_checks/_helpers/explain.rs → crates/sqlbuild-analysis/src/semantic_checks/_helpers/explanation/explain.rs

@@ -5,11 +5,11 @@ use std::sync::LazyLock;
 
 use regex::Regex;
 
-use crate::semantic_checks::_helpers::columns::{closest_column, ordered_columns};
-use crate::semantic_checks::_helpers::messages::{
+use crate::semantic_checks::_helpers::explanation::columns::{closest_column, ordered_columns};
+use crate::semantic_checks::_helpers::explanation::messages::{
     comparison_help, compiled, missing_column, pattern, semantic_help, sentence_message, type_words,
 };
-use crate::semantic_checks::_helpers::text::{
+use crate::semantic_checks::_helpers::sql_text::text::{
     LineIndex, ascii, last_newline_before, newlines_before, prefix,
 };
 use crate::semantic_checks::constants::{

crates/sqlbuild-analysis/src/semantic_checks/_helpers/messages.rs → crates/sqlbuild-analysis/src/semantic_checks/_helpers/explanation/messages.rs

@@ -4,7 +4,7 @@ use std::sync::LazyLock;
 
 use regex::{Captures, Regex};
 
-use crate::semantic_checks::_helpers::text::ascii;
+use crate::semantic_checks::_helpers::sql_text::text::ascii;
 use crate::semantic_checks::constants::{
     BIGQUERY_DIALECT, COMPARISON_CODE, CONTEXT_SUFFIX_PATTERN, DATE_TYPE, MISSING_PATTERN,
     QUOTED_PIECE_PATTERN, SEMANTIC_HELP, TIMESTAMP_TYPE, TYPE_WORDS_PATTERN,

crates/sqlbuild-analysis/src/semantic_checks/_helpers/explanation/mod.rs

@@ -0,0 +1,5 @@
+pub(crate) mod columns;
+pub(crate) mod completion;
+pub(crate) mod explain;
+pub(crate) mod messages;
+pub(crate) mod opt_outs;

crates/sqlbuild-analysis/src/semantic_checks/_helpers/opt_outs.rs → crates/sqlbuild-analysis/src/semantic_checks/_helpers/explanation/opt_outs.rs

@@ -4,7 +4,7 @@ use std::sync::LazyLock;
 
 use regex::Regex;
 
-use crate::semantic_checks::_helpers::messages::{compiled, pattern};
+use crate::semantic_checks::_helpers::explanation::messages::{compiled, pattern};
 use crate::semantic_checks::constants::{
     ADDITIONAL_HELP_SEPARATOR, FINDING_KINDS, PROJECT_CONFIG_FILENAME, REQUIRE_SQL_ANALYSIS_KEY,
     SETTING_SNIPPET_INDENT, SETTINGS_SECTION,

crates/sqlbuild-analysis/src/semantic_checks/_helpers/metadata_checks/argument_types.rs

@@ -0,0 +1,92 @@
+//! Python's `_polyglot_expression_type` for one function argument expression.
+
+use std::collections::HashMap;
+
+use polyglot_sql::{DataType, Expression};
+use serde_json::Value;
+
+use crate::semantic_checks::_helpers::sql_text::text::ascii;
+use crate::semantic_checks::constants::{
+    BOOLEAN_RESULT_KINDS, BOOLEAN_TYPE_NAME, COLUMN_KIND, CUSTOM_TYPE_NAME, DECIMAL_TYPE_NAME,
+    KNOWN_CAST_TYPE_NAMES, TIMESTAMP_KIND, TIMESTAMP_WITH_TIME_ZONE_TYPE_NAME, VARCHAR_DATA_TYPES,
+};
+use crate::semantic_checks::models::SemanticDeferral;
+
+/// The type Python infers for `expression`, or None.
+pub(crate) fn expression_type(
+    expression: &Expression,
+    return_types: &HashMap<String, String>,
+) -> Result<Option<String>, SemanticDeferral> {
+    let kind: &str = expression.variant_name();
+    let name: &str = expression.get_name();
+    let function_name: &str = if name.is_empty() { kind } else { name };
+    if kind != COLUMN_KIND
+        && let Some(declared) = return_types.get(&ascii(function_name)?.to_ascii_uppercase())
+    {
+        return Ok(Some(declared.clone()));
+    }
+    if BOOLEAN_RESULT_KINDS.contains(&kind) {
+        return Ok(Some(BOOLEAN_TYPE_NAME.to_owned()));
+    }
+    match expression {
+        Expression::Cast(cast) | Expression::TryCast(cast) => cast_type(&cast.to),
+        _ => Ok(None),
+    }
+}
+
+/// Python's reading of a cast's serialised `to` payload.
+fn cast_type(data_type: &DataType) -> Result<Option<String>, SemanticDeferral> {
+    let target: Value =
+        serde_json::to_value(data_type).map_err(|_| SemanticDeferral::NativeFailure)?;
+    let Some(raw_type) = target
+        .get("data_type")
+        .and_then(Value::as_str)
+        .filter(|raw| !raw.is_empty())
+    else {
+        return Ok(None);
+    };
+    let raw_type: &str = ascii(raw_type)?;
+    let custom_name: Option<&str> = target
+        .get("name")
+        .and_then(Value::as_str)
+        .filter(|name| !name.is_empty());
+    if raw_type.eq_ignore_ascii_case(CUSTOM_TYPE_NAME)
+        && let Some(name) = custom_name
+    {
+        return Ok(Some(ascii(name)?.to_ascii_uppercase()));
+    }
+    let lowered: String = raw_type.to_ascii_lowercase();
+    if lowered == TIMESTAMP_KIND && target.get("timezone") == Some(&Value::Bool(true)) {
+        return Ok(Some(TIMESTAMP_WITH_TIME_ZONE_TYPE_NAME.to_owned()));
+    }
+    let type_name: String = KNOWN_CAST_TYPE_NAMES
+        .iter()
+        .find(|(raw, _)| *raw == raw_type)
+        .map_or_else(
+            || raw_type.replace('_', " ").to_ascii_uppercase(),
+            |(_, known)| (*known).to_owned(),
+        );
+    if VARCHAR_DATA_TYPES.contains(&lowered.as_str())
+        && let Some(length) = integer(&target, "length")
+    {
+        return Ok(Some(format!("VARCHAR({length})")));
+    }
+    if type_name == DECIMAL_TYPE_NAME
+        && let Some(precision) = integer(&target, "precision")
+    {
+        return Ok(Some(match integer(&target, "scale") {
+            Some(scale) => format!("DECIMAL({precision}, {scale})"),
+            None => format!("DECIMAL({precision})"),
+        }));
+    }
+    Ok(Some(type_name))
+}
+
+/// Python's `isinstance(target.get(key), int)` value.
+fn integer(target: &Value, key: &str) -> Option<i128> {
+    let value: &Value = target.get(key)?;
+    value
+        .as_i64()
+        .map(i128::from)
+        .or_else(|| value.as_u64().map(i128::from))
+}