@malloy-publisher/server 0.0.235 → 0.0.236

package/src/constants.ts

@@ -1,105 +0,0 @@
-import os from "os";
-export const API_PREFIX = "/api/v0";
-export const README_NAME = "README.md";
-export const PUBLISHER_CONFIG_NAME = "publisher.config.json";
-export const PACKAGE_MANIFEST_NAME = "publisher.json";
-export const MODEL_FILE_SUFFIX = ".malloy";
-export const NOTEBOOK_FILE_SUFFIX = ".malloynb";
-
-/**
- * Normalize a package-relative model path so author-written `explores`
- * entries compare equal to the paths produced by `listPackageFiles`
- * (forward slashes, no leading "./"). Shared by every input channel — the
- * worker's on-disk parse and the API publish/update path — so a manifest
- * behaves the same whether it's loaded from disk or written via the API.
- */
-export function normalizeModelPath(p: string): string {
-   return p.replace(/\\/g, "/").replace(/^\.\//, "");
-}
-
-/**
- * Default row cap applied to Malloy model queries (the `runnable.run`
- * path used by `getQueryResults` and notebook cell execution) when
- * the user's query doesn't carry its own `LIMIT`. Override at startup
- * via `PUBLISHER_DEFAULT_QUERY_ROW_LIMIT`.
- *
- * This is *the default*, not a hard ceiling. A Malloy query that
- * carries `LIMIT 50000` overrides this. The hard ceiling is
- * {@link DEFAULT_MAX_QUERY_ROWS} (env-tuned via
- * `PUBLISHER_MAX_QUERY_ROWS`), which the model-query path now also
- * enforces — preventing a user `LIMIT 10000000` from blowing up the
- * process.
- *
- * Tuned conservatively (1000) because notebook UIs typically render
- * tabular previews; operators who need larger ad-hoc exports can
- * raise it.
- */
-export const DEFAULT_QUERY_ROW_LIMIT = 1000;
-/**
- * Maximum number of rows the ad-hoc connection SQL endpoints
- * (`/environments/.../connections/.../sqlQuery`) will return in a
- * single response before failing with HTTP 413. Override at startup
- * via `PUBLISHER_MAX_QUERY_ROWS`.
- *
- * Tuned high enough to handle typical interactive SQL exploration
- * without truncating, while still bounding memory: 100k narrow rows
- * is ~tens of MB of buffered result + a same-order JSON.stringify
- * copy, comfortably under any reasonable pod memory budget. Wide
- * rows (large strings, JSON blobs) are bounded separately by the
- * Step 2 byte budget.
- */
-export const DEFAULT_MAX_QUERY_ROWS = 100_000;
-/**
- * Maximum aggregate JSON-serialized byte size of an ad-hoc
- * connection SQL response before failing with HTTP 413. Override at
- * startup via `PUBLISHER_MAX_RESPONSE_BYTES`.
- *
- * This is the actual memory bound: row count alone is a poor proxy
- * for memory pressure because a single 10 MB JSON column blows past
- * the 100k-row cap's safe envelope. Enforced only when the
- * underlying connection implements `StreamingConnection` (Postgres,
- * DuckDB, ...) since byte counting requires iterating row-at-a-time
- * — on non-streaming connections the driver has already buffered
- * the whole result by the time we see it, so client-side byte
- * counting gains nothing.
- *
- * 50 MB picked as a middle ground: comfortably accommodates wide
- * exploratory results (50k rows × 1 KB, or 5k rows × 10 KB) while
- * keeping a single pod able to serve a handful of concurrent
- * requests under a typical 1-2 GB memory budget.
- */
-export const DEFAULT_MAX_RESPONSE_BYTES = 50_000_000;
-/**
- * Default wall-clock timeout (milliseconds) applied to a single query
- * — whether it's an ad-hoc connection SQL call or a Malloy model
- * query. Override at startup via `PUBLISHER_QUERY_TIMEOUT_MS`; `0`
- * disables the timeout entirely.
- *
- * 5 minutes is generous enough to let ad-hoc materializations and
- * larger interactive queries finish, while still well under the
- * 10-minute HTTP socket timeout so the publisher returns a clean
- * `504 Gateway Timeout` instead of letting the load balancer
- * disconnect mid-stream. Operators running long-tail analytical
- * jobs should raise this; operators running a high-RPS API surface
- * may want to lower it.
- */
-export const DEFAULT_QUERY_TIMEOUT_MS = 300_000;
-/**
- * Default upper bound on the number of inbound query requests a
- * single publisher pod will handle concurrently. Excess requests
- * are rejected with `503 Service Unavailable` so the upstream load
- * balancer can retry / shed load instead of letting them pile up
- * and stampede memory. Override at startup via
- * `PUBLISHER_MAX_CONCURRENT_QUERIES`; `0` disables the limit.
- *
- * 32 picked as a safe out-of-the-box value: it bounds aggregate
- * memory usage at ~32× the per-query byte cap
- * ({@link DEFAULT_MAX_RESPONSE_BYTES}, i.e. ~1.5 GB at the default
- * 50 MB) while leaving substantial headroom for the API surface,
- * notebook execution, and background materializations. Pods sized
- * larger than the default should raise it; pods running on tight
- * memory budgets should lower it.
- */
-export const DEFAULT_MAX_CONCURRENT_QUERIES = 32;
-export const TEMP_DIR_PATH = os.tmpdir();
-export const PUBLISHER_DATA_DIR = "publisher_data";

package/src/controller/compile.controller.ts

@@ -1,40 +0,0 @@
-import type { GivenValue, LogMessage } from "@malloydata/malloy";
-import { EnvironmentStore } from "../service/environment_store";
-
-export class CompileController {
-   private environmentStore: EnvironmentStore;
-
-   constructor(environmentStore: EnvironmentStore) {
-      this.environmentStore = environmentStore;
-   }
-
-   public async compile(
-      environmentName: string,
-      packageName: string,
-      modelName: string,
-      source: string,
-      includeSql: boolean = false,
-      givens?: Record<string, GivenValue>,
-   ): Promise<{ status: string; problems: LogMessage[]; sql?: string }> {
-      const environment = await this.environmentStore.getEnvironment(
-         environmentName,
-         false,
-      );
-      const { problems, sql } = await environment.compileSource(
-         packageName,
-         modelName,
-         source,
-         includeSql,
-         givens,
-      );
-
-      // Determine overall status based on presence of errors
-      const hasErrors = problems.some((p) => p.severity === "error");
-
-      return {
-         status: hasErrors ? "error" : "success",
-         problems: problems,
-         ...(sql !== undefined && { sql }),
-      };
-   }
-}