refactor: port clone, load, check, test, seed and diff commands, rule caches and state helpers to the native extension

Cargo.toml

@@ -2,6 +2,7 @@
 members = [
     "crates/sqlbuild-core",
     "crates/sqlbuild-cursor",
+    "crates/sqlbuild-state",
     "crates/sqlbuild-pysource",
     "crates/sqlbuild-cache",
     "crates/sqlbuild-sqltext",
@@ -26,6 +27,7 @@ resolver = "2"
 order = [
     "sqlbuild-core",
     "sqlbuild-cursor",
+    "sqlbuild-state",
     "sqlbuild-pysource",
     "sqlbuild-cache",
     "sqlbuild-sqltext",
@@ -104,6 +106,7 @@ sqlbuild-model-config = { path = "crates/sqlbuild-model-config" }
 sqlbuild-pysource = { path = "crates/sqlbuild-pysource" }
 sqlbuild-scopes = { path = "crates/sqlbuild-scopes" }
 sqlbuild-sqltext = { path = "crates/sqlbuild-sqltext" }
+sqlbuild-state = { path = "crates/sqlbuild-state" }
 stacker = "0.1"
 sqlparser = { version = "0.58", features = ["visitor"] }
 syn = { version = "2", features = ["full", "visit", "extra-traits"] }

crates/README.md

@@ -8,6 +8,7 @@ that can be tested and benchmarked with `cargo test` alone.
 |---|---|
 | `sqlbuild-core` | Constants shared by several crates, the panic boundary that turns an unwinding native failure into a named compiler error, passive identity, diagnostic and SQL value types, text positions, and a JSON emitter whose output matches Python's `json.dumps` and `orjson` byte for byte. |
 | `sqlbuild-cursor` | Microbatch cursor algebra: typed cursor values, grain flooring and stepping, half-open interval clamping, merging and batch splitting, timestamp text normalization and calendar durations, with Python `datetime` results and errors. Zones other than fixed offsets are resolved through a caller-supplied lookup. |
+| `sqlbuild-state` | Append-only warehouse state rows: content-addressed migration event IDs, typed event SQL literals and statements, and source freshness data-version normalization, hashing, equivalence and age policy, byte-compatible with the rows Python wrote. |
 | `sqlbuild-pysource` | Static reading of project Python source with the ruff parser: macro module structure, literal SQL arguments, definition locations and string-literal edits. It never imports or runs user code; CPython stays the authority on whether a file is valid Python. |
 | `sqlbuild-cache` | The shared native store every native cache uses (one atomically replaced, checksummed file per cache kind, keyed by content digests) and the one fingerprinting scheme for cache keys: the content digest and the project file fingerprint. |
 | `sqlbuild-sqltext` | Lexical SQL text without polyglot: comment, quote and parenthesis scanning, the rules quote policy, model header tokenization and matching, static variable substitution, static reference extraction, and the general reference scan with table-function call arguments under each adapter's lexical rules. |
@@ -31,7 +32,7 @@ Dependencies point one way, from the top of this graph to the bottom:
 sqlbuild-python
   -> sqlbuild-refactor -> sqlbuild-rules -> sqlbuild-analysis -> sqlbuild-attachments
   -> sqlbuild-model-config -> sqlbuild-scopes -> sqlbuild-render -> sqlbuild-discovery -> sqlbuild-config
-  -> sqlbuild-sqltext -> sqlbuild-cache -> sqlbuild-pysource -> sqlbuild-cursor -> sqlbuild-core
+  -> sqlbuild-sqltext -> sqlbuild-cache -> sqlbuild-pysource -> sqlbuild-state -> sqlbuild-cursor -> sqlbuild-core
 sqlbuild-cli -> sqlbuild-cache, sqlbuild-core
 ```
 
@@ -56,6 +57,21 @@ The same check fails when the workspace version, the `pyproject.toml` version an
 `fensu check` applies the Rust structure rules inside each crate: other crates use items through
 `main/` entry points, `models.rs`, `types.rs` and `constants.rs`, not through `_helpers/`.
 
+## Python module paths
+
+`sqlbuild-python` and `sqlbuild-cli` import Python by module path strings. The repository test
+`tests/unit/scripts/native_module_paths` (`scripts/native_module_paths`) keeps them honest:
+
+- every `"sqlbuild...."` literal must be a complete dotted path that imports as a module or
+  resolves as a module attribute; constants whose name contains `LOGGER` hold `logging`
+  logger names and are not imported;
+- attributes are checked where a literal module, or a module constant defined in the same file,
+  is passed with a literal name to an import helper (`imported(py, MODULE, "name")`), to
+  `import(MODULE)?.getattr("name")`, or written as a `("sqlbuild.module", "name")` pair;
+- module paths are never built at runtime: a `format!` or `concat!` template may not start with
+  `sqlbuild.` or with a `{..PACKAGE..}.`/`{..MODULE..}.` constant. Write the complete module
+  path as a constant instead.
+
 ## Compiler
 
 The compiler runs discovery, rendering and analysis natively. Generic compiler stages do not

crates/sqlbuild-cli/src/diff_report/_helpers/evidence.rs

@@ -0,0 +1,110 @@
+//! Bounded, difference-aware rendering of example values and keys.
+
+use sqlbuild_cache::digest::main::sha256_hex::sha256_hex;
+
+use crate::diff_report::_helpers::numbers::grouped;
+use crate::diff_report::constants::MAX_RENDERED_KEY_LENGTH;
+use crate::diff_report::models::{DiffExampleOptions, RenderedValue};
+
+fn first_difference(left: &[char], right: &[char]) -> Option<usize> {
+    if let Some(index) = left.iter().zip(right).position(|(a, b)| a != b) {
+        return Some(index);
+    }
+    (left.len() != right.len()).then(|| left.len().min(right.len()))
+}
+
+fn render_value(
+    text: &[char],
+    first_difference: Option<usize>,
+    options: DiffExampleOptions,
+) -> RenderedValue {
+    let original_length: usize = text.len();
+    let full: String = text.iter().collect();
+    if options.suppress_values {
+        return RenderedValue {
+            text: None,
+            original_length,
+            truncated: false,
+            suppressed: true,
+            first_difference,
+        };
+    }
+    let Some(limit) = options
+        .max_value_length
+        .filter(|limit| original_length > *limit)
+    else {
+        return RenderedValue {
+            text: Some(full),
+            original_length,
+            truncated: false,
+            suppressed: false,
+            first_difference,
+        };
+    };
+    let difference: usize = first_difference.unwrap_or(0);
+    let context_before: usize = (limit / 3).min(limit.saturating_sub(1));
+    let start: usize = difference
+        .saturating_sub(context_before)
+        .min(original_length - limit);
+    let end: usize = start + limit;
+    let fragment: String = text[start..end].iter().collect();
+    RenderedValue {
+        text: Some(format!(
+            "{}{fragment}{}",
+            if start > 0 { "…" } else { "" },
+            if end < original_length { "…" } else { "" }
+        )),
+        original_length,
+        truncated: true,
+        suppressed: false,
+        first_difference,
+    }
+}
+
+/// Two unequal values rendered around their first differing character.
+pub(crate) fn render_example_pair(
+    left: &str,
+    right: &str,
+    options: DiffExampleOptions,
+) -> (RenderedValue, RenderedValue) {
+    let left: Vec<char> = left.chars().collect();
+    let right: Vec<char> = right.chars().collect();
+    let difference: Option<usize> = first_difference(&left, &right);
+    (
+        render_value(&left, difference, options),
+        render_value(&right, difference, options),
+    )
+}
+
+/// One bounded value for terminal output.
+pub(crate) fn display_example_value(value: &RenderedValue) -> String {
+    if value.suppressed {
+        return format!(
+            "<value suppressed; {} chars>",
+            grouped(value.original_length as i64)
+        );
+    }
+    let rendered: &str = value.text.as_deref().unwrap_or_default();
+    if value.truncated {
+        return format!(
+            "{rendered} [truncated; {} chars]",
+            grouped(value.original_length as i64)
+        );
+    }
+    rendered.to_owned()
+}
+
+/// Ordinary keys complete; exceptionally long keys shortened and identified by digest.
+pub(crate) fn render_key_value(text: &str) -> String {
+    let length: usize = text.chars().count();
+    if length <= MAX_RENDERED_KEY_LENGTH {
+        return text.to_owned();
+    }
+    let digest: String = sha256_hex(text.as_bytes());
+    let prefix: String = text.chars().take(MAX_RENDERED_KEY_LENGTH).collect();
+    format!(
+        "{prefix}… [truncated; {} chars; sha256:{}]",
+        grouped(length as i64),
+        &digest[..12]
+    )
+}

crates/sqlbuild-cli/src/diff_report/_helpers/findings.rs

@@ -0,0 +1,35 @@
+//! Whether a diff found differences, and the resulting outcome.
+
+use crate::diff_report::constants::{FINDINGS_OUTCOME, PASS_OUTCOME};
+use crate::diff_report::models::DiffModelFacts;
+
+pub(crate) fn has_schema_differences(model: &DiffModelFacts) -> bool {
+    !(model.added_columns.is_empty()
+        && model.removed_columns.is_empty()
+        && model.type_changes.is_empty())
+}
+
+/// Whether any model has schema or row differences.
+pub(crate) fn has_findings(models: &[DiffModelFacts]) -> bool {
+    models.iter().any(|model| {
+        has_schema_differences(model)
+            || model.rows.as_ref().is_some_and(|rows| {
+                rows.unequal_count != 0 || rows.left_only_count != 0 || rows.right_only_count != 0
+            })
+    })
+}
+
+/// The explicit outcome, or `findings`/`pass` from the results.
+pub(crate) fn resolved_outcome(models: &[DiffModelFacts], outcome: Option<&str>) -> String {
+    outcome
+        .filter(|outcome| !outcome.is_empty())
+        .map(str::to_owned)
+        .unwrap_or_else(|| {
+            if has_findings(models) {
+                FINDINGS_OUTCOME
+            } else {
+                PASS_OUTCOME
+            }
+            .to_owned()
+        })
+}