fix+refactor: soft keywords as identifiers, full test wiring, ORC crash docs

- parser: clause keywords (header, format, status, user, csv, ...) now work as identifiers everywhere; IMPORT/EXPORT accept FORMAT csv/HEADER true - nimble test + CI run all 13 test suites (650 checks green) - ExecutionContext.registry is now {.cursor.} (breaks registry<->ctx cycle) - ORC crash reproduced and bisected (tests/orc_repro.py); ARC stays the MM
2026-07-30 13:11:28 +03:00
parent ed5a71913c
commit 2d09edd9f7
9 changed files with 952 additions and 94 deletions
@@ -250,3 +250,74 @@ suite "Bug fixes — IN list, nkPath exprToSql, multi-table joins":
    check rmax.success
    check rmax.rows.len == 1
    check valueToString(rmax.rows[0]["m"]) == "30"
+
+suite "Bug fixes — keyword 'header' usable as column name":
+
+  test "CREATE TABLE / INSERT / SELECT with column named 'header'":
+    var ctx = setupCtx()
+    defer: teardown(ctx)
+    # nimforum schema uses a column named 'header' (tkHeader is the CSV IMPORT keyword)
+    let c = executeQuery(ctx, parse("CREATE TABLE post (id INTEGER PRIMARY KEY, header TEXT, content TEXT)"))
+    check c.success
+    let i = executeQuery(ctx, parse("INSERT INTO post (id, header, content) VALUES (1, 'Hello', 'World')"))
+    check i.success
+    let r = executeQuery(ctx, parse("SELECT header FROM post WHERE id = 1"))
+    check r.success
+    check r.rows.len == 1
+    check valueToString(r.rows[0]["header"]) == "Hello"
+    let rw = executeQuery(ctx, parse("SELECT id FROM post WHERE header = 'Hello'"))
+    check rw.success
+    check rw.rows.len == 1
+
+  test "IMPORT ... HEADER clause still parses after soft-keyword change":
+    let ast = parse("IMPORT FROM 'data.csv' INTO post HEADER no")
+    check ast.stmts.len == 1
+    check ast.stmts[0].kind == nkImportFrom
+    check ast.stmts[0].impHasHeader == false
+    let ast2 = parse("EXPORT TO 'out.csv' FROM post HEADER yes")
+    check ast2.stmts.len == 1
+    check ast2.stmts[0].kind == nkExportTo
+    check ast2.stmts[0].expIncludeHeader == true
+
+suite "Bug fixes — clause keywords usable as identifiers":
+
+  test "columns named after clause keywords (format, status, user, ...)":
+    var ctx = setupCtx()
+    defer: teardown(ctx)
+    let c = executeQuery(ctx, parse("""
+      CREATE TABLE kw (id INTEGER PRIMARY KEY, format TEXT, status TEXT, user TEXT,
+                       batch INTEGER, csv TEXT, ndjson TEXT, delimiter TEXT,
+                       migration TEXT, apply TEXT, up TEXT, down TEXT, dryrun TEXT,
+                       policy TEXT, enable TEXT, disable TEXT, recover TEXT,
+                       before TEXT, after TEXT, instead TEXT, of TEXT)
+    """))
+    check c.success
+    let i = executeQuery(ctx, parse(
+      "INSERT INTO kw (id, format, status, user, batch, of) VALUES (1, 'csv', 'active', 'admin', 7, 'x')"))
+    check i.success
+    let u = executeQuery(ctx, parse("UPDATE kw SET status = 'done' WHERE id = 1"))
+    check u.success
+    let r = executeQuery(ctx, parse("SELECT format, status, user, batch, of FROM kw WHERE id = 1"))
+    check r.success
+    check r.rows.len == 1
+    check valueToString(r.rows[0]["format"]) == "csv"
+    check valueToString(r.rows[0]["status"]) == "done"
+    check valueToString(r.rows[0]["user"]) == "admin"
+    check valueToString(r.rows[0]["batch"]) == "7"
+    let rq = executeQuery(ctx, parse("SELECT kw.status FROM kw WHERE kw.status = 'done'"))
+    check rq.success
+    check rq.rows.len == 1
+
+  test "IMPORT/EXPORT accept keyword values: FORMAT csv/ndjson/json, HEADER true/false":
+    let a1 = parse("IMPORT FROM 'd.csv' INTO t FORMAT csv HEADER true")
+    check a1.stmts[0].impFormat == "csv"
+    check a1.stmts[0].impHasHeader == true
+    let a2 = parse("IMPORT FROM 'd.csv' INTO t FORMAT ndjson HEADER false")
+    check a2.stmts[0].impFormat == "ndjson"
+    check a2.stmts[0].impHasHeader == false
+    let a3 = parse("EXPORT TO 'o.csv' FROM t FORMAT json HEADER true")
+    check a3.stmts[0].expFormat == "json"
+    check a3.stmts[0].expIncludeHeader == true
+    let a4 = parse("EXPORT TO 'o.csv' FROM t FORMAT csv DELIMITER ';' HEADER false")
+    check a4.stmts[0].expFormat == "csv"
+    check a4.stmts[0].expIncludeHeader == false
@@ -0,0 +1,57 @@
+#!/usr/bin/env python3
+"""Reproducer for the ORC cycle-collector crash (markGray/trace SIGSEGV).
+
+Usage:
+  1. Build the server with ORC:
+       nim c -d:ssl --threads:on --path:src --mm:orc -o:/tmp/baradadb_orc src/baradadb.nim
+  2. Start it:
+       BARADB_PORT=39472 BARADB_DATA_DIR=/tmp/baradb_orc_data BARADB_LOG_LEVEL=error /tmp/baradadb_orc
+  3. Run this script (from the repo root):
+       python3 tests/orc_repro.py
+
+Expected (as of 2026-07-30, Nim 2.2.10): the server dies with
+  orc.nim markGray -> trace -> SIGSEGV
+triggered from core/server.nim handleClient, and this script fails with
+ConnectionResetError. Under ARC (the default in nim.cfg) the same load passes.
+
+Bisect results: 200 pings + 200 SELECTs over TCP are fine; the crash lands
+somewhere between 20 and 500 sequential INSERTs (INSERT path only).
+Failed root-cause attempts: callback cycle breaks (ed5a719), {.cursor.} on
+ExecutionContext.registry (the registry<->ctx cycle), guarding ctx.onChange
+against zero WS subscribers. Conclusion: deep ORC+async issue (possibly
+upstream Nim), not a single app-level cycle. ARC remains the supported MM.
+"""
+
+import asyncio
+import sys
+
+sys.path.insert(0, "clients/python")
+from baradb import Client
+
+PORT = 39472
+
+
+async def worker(w: int) -> None:
+    base = 1000 + w * 100
+    async with Client("127.0.0.1", PORT) as c:
+        for i in range(100):
+            await c.query(f"INSERT INTO orc_stress (id, val) VALUES ({base + i}, 'v{base + i}')")
+
+
+async def main() -> None:
+    async with Client("127.0.0.1", PORT) as c:
+        await c.query("CREATE TABLE orc_stress (id INT PRIMARY KEY, val STRING)")
+        # Original report: SIGSEGV after ~20 sequential INSERTs under async load
+        for i in range(1000):
+            await c.query(f"INSERT INTO orc_stress (id, val) VALUES ({i}, 'x{i}')")
+        r = await c.query("SELECT COUNT(*) AS n FROM orc_stress")
+        print("after 1000 sequential INSERTs:", r.rows)
+    await asyncio.gather(*[worker(w) for w in range(10)])
+    async with Client("127.0.0.1", PORT) as c:
+        r = await c.query("SELECT COUNT(*) AS n FROM orc_stress")
+        print("final count:", r.rows)
+        print("ping:", await c.ping())
+    print("ORC STRESS OK")
+
+
+asyncio.run(main())