feat!: name built-in Rules SQLCLEAN and own the host contract

Built-in Rule codes are SQLCLEANnnn instead of SQCnnn, custom Rule codes
start with XSQLCLEAN, and the parse error is SQLCLEAN999.

The embedding API no longer carries SQLBuild names. Hosts write the
SQLClean placeholder prefixes exported from constants, describe a test
or macro wrapper with one fixture_harness request object instead of
allows_ceremonial_select and allows_empty_fixture_star, and derive
relation_keys themselves: DeclaredRelation and declared_relation_keys
are removed.

BREAKING CHANGE: SQC and XSQC codes are now SQLCLEAN and XSQLCLEAN, and
the lint request replaces the two harness flags with fixture_harness.

Author: kvlonge

.github/demos/check-fix.gif

@@ -67,7 +67,7 @@ SQLClean in other configurations, on the same files:
 
 | SQLClean command | Wall time |
 |---|---:|
-| `sqlclean check --select SQC` (every built-in Rule) | 1.33 s |
+| `sqlclean check --select SQLCLEAN` (every built-in Rule) | 1.33 s |
 | `sqlclean check` plus three custom Python Rules | 2.27 s |
 | `sqlclean check` rerun with nothing changed (result cache) | 0.23 s |
 
@@ -114,7 +114,7 @@ def lowered(operand: ast.Expression) -> bool:
 
 
 @rule(
-    code="XSQCTEAM001",
+    code="XSQLCLEANTEAM001",
     message="Email comparisons must be case-insensitive",
     remediation="Compare LOWER(email) with LOWER(...) instead.",
     enabled_by_default=True,

README.md

@@ -34,7 +34,7 @@ and peak memory with `/usr/bin/time`, and every lint command writes JSON finding
 | Variant | Command |
 |---|---|
 | `sqlclean` | `sqlclean check --output-format json` (13 default Rules) |
-| `sqlclean_all` | `sqlclean check --select SQC --output-format json` (all 45 Rules) |
+| `sqlclean_all` | `sqlclean check --select SQLCLEAN --output-format json` (all 45 Rules) |
 | `sqlclean_custom` | `sqlclean check --config custom.toml --output-format json` (default Rules and the three Python Rules in `rules/`) |
 | `sqlclean_format` | `sqlclean format --check` |
 | `sqruff` | `sqruff lint -f json .` |

benchmarks/linters/README.md

@@ -34,7 +34,7 @@ trap 'rm -rf "$WORK"' EXIT
 
 declare -A CMD
 CMD[sqlclean]="$SQLCLEAN check --output-format json"
-CMD[sqlclean_all]="$SQLCLEAN check --select SQC --output-format json"
+CMD[sqlclean_all]="$SQLCLEAN check --select SQLCLEAN --output-format json"
 # The default Rules plus the three Python Rules in rules/, copied into the work directory.
 CMD[sqlclean_custom]="$SQLCLEAN check --config custom.toml --output-format json"
 # sqruff's and SQLFluff's default rules include layout; SQLClean checks layout with its formatter.

benchmarks/linters/bench.sh

@@ -8,7 +8,7 @@
 
 
 @rule(
-    code="XSQCLAYER001",
+    code="XSQLCLEANLAYER001",
     message="Mart reads a raw source",
     remediation="Read the staging model instead.",
     enabled_by_default=True,
@@ -26,7 +26,7 @@ def marts_read_staging(*, statement: Statement, ctx: RuleContext) -> list[Findin
 
 
 @rule(
-    code="XSQCLAYER002",
+    code="XSQLCLEANLAYER002",
     message="Staging model reads another model",
     remediation="Staging models read only raw sources and seeds.",
     enabled_by_default=True,
@@ -46,7 +46,7 @@ def staging_reads_sources(*, statement: Statement, ctx: RuleContext) -> list[Fin
 
 
 @rule(
-    code="XSQCDET001",
+    code="XSQLCLEANDET001",
     message="Model calls a nondeterministic function",
     remediation="Pass the time or seed in as a column so reruns give the same rows.",
     enabled_by_default=True,

benchmarks/linters/rules/layers.py

@@ -40,7 +40,7 @@ def count_findings(variant: str, out: str) -> tuple[int, int]:
     data = json.loads(out)
     if variant.startswith("sqlclean"):
         codes = [finding["code"] for finding in data["findings"]]
-        return len(codes), codes.count("SQC999")
+        return len(codes), codes.count("SQLCLEAN999")
     if variant == "sqruff":
         codes = [finding.get("code") for findings in data.values() for finding in findings]
         return len(codes), sum(1 for code in codes if code in (None, "PRS"))

benchmarks/linters/summarize.py

@@ -1,6 +1,6 @@
 ---
 name: sqlclean
-description: ALWAYS load this skill when linting, fixing or formatting SQL with SQLClean, editing sqlclean.toml, writing or testing custom Rules (Python @rule functions with XSQC codes), recording Rule exceptions, or reading SQLClean findings (SQC codes, SQC999 parse errors) in CI, JSON or SARIF output.
+description: ALWAYS load this skill when linting, fixing or formatting SQL with SQLClean, editing sqlclean.toml, writing or testing custom Rules (Python @rule functions with XSQLCLEAN codes), recording Rule exceptions, or reading SQLClean findings (SQLCLEAN codes, SQLCLEAN999 parse errors) in CI, JSON or SARIF output.
 ---
 <!-- generated-by: sqlclean skills -->
 
@@ -17,7 +17,7 @@ engine extracts from each statement.
    `[rules]`). Paths in it are relative to its directory.
 2. **Check.** `sqlclean check [PATHS]` exits 0 when clean, 1 with findings, 2 on errors. Use
    `--output-format json` when you need to read findings programmatically.
-3. **Explain.** `sqlclean rules SQC001` shows a Rule's message, fix safety and remediation.
+3. **Explain.** `sqlclean rules SQLCLEAN001` shows a Rule's message, fix safety and remediation.
 4. **Fix.** `sqlclean fix --diff` previews safe fixes plus formatting; `sqlclean fix` writes them.
    Fixes marked unsafe can change query results, such as `= NULL` becoming `IS NULL`. Never pass
    `--unsafe-fixes` without the user's say-so.
@@ -31,16 +31,16 @@ engine extracts from each statement.
   `[[rules.rule_exceptions]]` entry with the Rule, path and a `reason`. Use
   `[[rules.rule_ignores]]` only for whole paths a Rule does not apply to.
 - An exception that no longer matches a finding is reported as stale; remove it.
-- `SQC999` is a parse error. It cannot be ignored: fix the SQL or set the right `dialect`,
+- `SQLCLEAN999` is a parse error. It cannot be ignored: fix the SQL or set the right `dialect`,
   globally or per path under `[[dialects]]`.
-- `--select` and `--ignore` take codes or prefixes such as `SQC001` or `SQC`.
+- `--select` and `--ignore` take codes or prefixes such as `SQLCLEAN001` or `SQLCLEAN`.
 - Results are cached in `.sqlclean/cache/`. Use `--no-cache` only when debugging a custom Rule
   that imports helpers from outside its `[rules] custom` paths.
 
 ## Custom Rules
 
 - A custom Rule is a function decorated with `@rule` from `sqlclean`, in a file under a path
-  listed in `[rules] custom`. Its code is `XSQC`, optional capital letters, then three digits.
+  listed in `[rules] custom`. Its code is `XSQLCLEAN`, optional capital letters, then three digits.
 - The first keyword parameter names the subject: `statement: Statement`, `file: SqlFile` or
   `repository: Repository`. The second is `ctx: RuleContext`.
 - Return `ctx.finding(...)` values. Fixes are lists of `ctx.replace`, `ctx.insert_before`,

crates/sqlclean-cli/skill/SKILL.md

@@ -14,7 +14,7 @@ SQLFluff ran once.
 | Command | Wall time | CPU time | Peak memory | vs SQLClean defaults |
 |---|---:|---:|---:|---:|
 | `sqlclean check` (13 default Rules) | **0.81 s** | 2.5 s | 48 MiB | 1× |
-| `sqlclean check --select SQC` (all 45 Rules) | **1.33 s** | 3.8 s | 140 MiB | 1.6× |
+| `sqlclean check --select SQLCLEAN` (all 45 Rules) | **1.33 s** | 3.8 s | 140 MiB | 1.6× |
 | `sqlclean check` with three [custom Rules](#custom-rules) | **2.27 s** | 6.9 s | 182 MiB | 2.8× |
 | `sqlclean format --check` | **3.03 s** | 11.5 s | 49 MiB | 3.7× |
 | `sqruff lint` | 41.0 s | 114 s | 235 MiB | 51× |

crates/sqlclean-cli/skill/references/docs/benchmarks.md

@@ -29,7 +29,7 @@ sqlclean check [OPTIONS] [PATHS]...
 ```bash
 sqlclean check                              # every matching file under the configuration root
 sqlclean check models/orders.sql reports/   # only these files and directories
-sqlclean check --select SQC001,SQC004       # only these Rules
+sqlclean check --select SQLCLEAN001,SQLCLEAN004       # only these Rules
 sqlclean check --output-format sarif > sqlclean.sarif
 ```
 
@@ -50,28 +50,28 @@ when `include` does not match it, unless `exclude` does. See
 | `--config <PATH>` | Use this configuration file instead of the nearest `sqlclean.toml` |
 
 `--select` and `--ignore` take a comma-separated list, or the flag repeated. A value is an exact
-code such as `SQC001` or a prefix such as `SQC` or `XSQCNAME`. A value that matches no Rule is an
+code such as `SQLCLEAN001` or a prefix such as `SQLCLEAN` or `XSQLCLEANNAME`. A value that matches no Rule is an
 error. See [Selecting Rules](../rules/overview.md#selecting-rules).
 
 ## Text output
 
 ```text
-error[SQC004]: Row selection is nondeterministic
+error[SQLCLEAN004]: Row selection is nondeterministic
   --> models/customers.sql:6:1
    |
  6 | LIMIT 10;
    | ^^^^^
   = help: Add ORDER BY with a deterministic tie-breaker before LIMIT or OFFSET.
 
-error[SQC001]: Comparison with NULL is never true
+error[SQLCLEAN001]: Comparison with NULL is never true
   --> models/orders.sql:8:22
    |
  8 | WHERE o.cancelled_at = NULL
    |                      ^
   = help: Use IS NULL or IS NOT NULL when testing for NULL.
   = fix: available with `sqlclean fix --unsafe-fixes`
 
-error: stale rule exception: SQC004 in models/orders.sql matches no finding; remove it from sqlclean.toml
+error: stale rule exception: SQLCLEAN004 in models/orders.sql matches no finding; remove it from sqlclean.toml
 
 Found 3 problems (1 fixable with `sqlclean fix --unsafe-fixes`).
 ```
@@ -92,7 +92,7 @@ and count characters, not bytes.
 {
   "findings": [
     {
-      "code": "SQC004",
+      "code": "SQLCLEAN004",
       "path": "models/customers.sql",
       "line": 6,
       "column": 1,
@@ -103,7 +103,7 @@ and count characters, not bytes.
       "fix": null
     },
     {
-      "code": "SQC001",
+      "code": "SQLCLEAN001",
       "path": "models/orders.sql",
       "line": 8,
       "column": 22,
@@ -116,7 +116,7 @@ and count characters, not bytes.
   ],
   "stale_exceptions": [
     {
-      "rule": "SQC004",
+      "rule": "SQLCLEAN004",
       "path": "models/orders.sql",
       "relation": null
     }
@@ -140,7 +140,7 @@ text:
 
 ```json
 {
-  "ruleId": "SQC001",
+  "ruleId": "SQLCLEAN001",
   "level": "error",
   "message": {
     "text": "Comparison with NULL is never true. Use IS NULL or IS NOT NULL when testing for NULL."

crates/sqlclean-cli/skill/references/docs/cli/check.md

@@ -36,14 +36,14 @@ Files are chosen as for [`sqlclean check`](check.md).
 | `--no-cache` | Evaluate every file instead of reusing results for unchanged inputs |
 | `--config <PATH>` | Use this configuration file instead of the nearest `sqlclean.toml` |
 
-Only fixes of selected Rules apply. `sqlclean fix --select SQC005` removes unused CTEs and
+Only fixes of selected Rules apply. `sqlclean fix --select SQLCLEAN005` removes unused CTEs and
 formats the files, and leaves every other finding alone.
 
 ## Output
 
 ```text
 Fixed 1 problem(s) and changed 1 file(s).
-error[SQC001]: Comparison with NULL is never true
+error[SQLCLEAN001]: Comparison with NULL is never true
   --> models/open_orders.sql:7:24
    |
  7 |     WHERE cancelled_at = NULL
@@ -88,7 +88,7 @@ findings:
   formatter cannot format is reported as `<path>: could not format: <reason>` and keeps its fixes.
 
 A file that does not parse gets no fixes and is reported as
-[`SQC999`](../rules/overview.md#parse-errors-sqc999).
+[`SQLCLEAN999`](../rules/overview.md#parse-errors-sqlclean999).
 
 ## Exit codes