Improve Gremlin schema sync performance

Changelog.md

@@ -6,6 +6,9 @@
 
 - Search in openCypher will now execute a single request when searching across
   all labels (<https://github.com/aws/graph-explorer/pull/493>)
+- Gremlin schema sync will be much faster on larger databases, thanks to
+  @dsaban-lightricks for his great suggestion in issue #225
+  (<https://github.com/aws/graph-explorer/pull/498>)
 
 **Bug Fixes and Minor Changes**
 

packages/graph-explorer/src/connector/gremlin/gremlinExplorer.ts

@@ -8,7 +8,7 @@ import { fetchDatabaseRequest } from "../fetchDatabaseRequest";
 import { GraphSummary } from "./types";
 import { v4 } from "uuid";
 import { Explorer } from "../useGEFetchTypes";
-import { env } from "../../utils";
+import { logger } from "../../utils";
 
 function _gremlinFetch(connection: ConnectionConfig, options: any) {
   return async (queryTemplate: string) => {
@@ -30,41 +30,53 @@ function _gremlinFetch(connection: ConnectionConfig, options: any) {
   };
 }
 
+async function fetchSummary(
+  connection: ConnectionConfig,
+  options: RequestInit
+) {
+  try {
+    const response = await fetchDatabaseRequest(
+      connection,
+      `${connection.url}/pg/statistics/summary?mode=detailed`,
+      {
+        method: "GET",
+        ...options,
+      }
+    );
+    return response.payload.graphSummary as GraphSummary;
+  } catch (error) {
+    logger.error(
+      "[Gremlin Explorer] Failed to gather summary statistics",
+      error
+    );
+  }
+}
+
 export function createGremlinExplorer(connection: ConnectionConfig): Explorer {
   return {
     connection: connection,
     async fetchSchema(options) {
-      let summary;
-      try {
-        const response = await fetchDatabaseRequest(
-          connection,
-          `${connection.url}/pg/statistics/summary?mode=detailed`,
-          {
-            method: "GET",
-            ...options,
-          }
-        );
-        summary = (response.payload.graphSummary as GraphSummary) || undefined;
-      } catch (e) {
-        if (env.DEV) {
-          console.error("[Summary API]", e);
-        }
-      }
+      logger.log("[Gremlin Explorer] Fetching schema...");
+      const summary = await fetchSummary(connection, options);
       return fetchSchema(_gremlinFetch(connection, options), summary);
     },
     async fetchVertexCountsByType(req, options) {
+      logger.log("[Gremlin Explorer] Fetching vertex counts by type...");
       return fetchVertexTypeCounts(_gremlinFetch(connection, options), req);
     },
     async fetchNeighbors(req, options) {
+      logger.log("[Gremlin Explorer] Fetching neighbors...");
       return fetchNeighbors(_gremlinFetch(connection, options), req);
     },
     async fetchNeighborsCount(req, options) {
+      logger.log("[Gremlin Explorer] Fetching neighbors count...");
       return fetchNeighborsCount(_gremlinFetch(connection, options), req);
     },
     async keywordSearch(req, options) {
       options ??= {};
       options.queryId = v4();
 
+      logger.log("[Gremlin Explorer] Fetching keyword search...");
       return keywordSearch(_gremlinFetch(connection, options), req);
     },
   };

packages/graph-explorer/src/connector/gremlin/queries/fetchSchema.ts

@@ -1,4 +1,4 @@
-import { sanitizeText } from "../../../utils";
+import { logger, sanitizeText } from "../../../utils";
 import type { SchemaResponse } from "../../useGEFetchTypes";
 import edgeLabelsTemplate from "../templates/edgeLabelsTemplate";
 import edgesSchemaTemplate from "../templates/edgesSchemaTemplate";
@@ -87,6 +87,7 @@ const fetchVertexLabels = async (
   gremlinFetch: GremlinFetch
 ): Promise<Record<string, number>> => {
   const labelsTemplate = vertexLabelsTemplate();
+  logger.log("[Gremlin Explorer] Fetching vertex labels with counts...");
   const data = await gremlinFetch<RawVertexLabelsResponse>(labelsTemplate);
 
   const values = data.result.data["@value"][0]["@value"];
@@ -121,6 +122,7 @@ const fetchVerticesAttributes = async (
     types: labels,
   });
 
+  logger.log("[Gremlin Explorer] Fetching vertices attributes...");
   const response =
     await gremlinFetch<RawVerticesSchemaResponse>(verticesTemplate);
   const verticesSchemas = response.result.data["@value"][0]["@value"];
@@ -163,6 +165,7 @@ const fetchEdgeLabels = async (
   gremlinFetch: GremlinFetch
 ): Promise<Record<string, number>> => {
   const labelsTemplate = edgeLabelsTemplate();
+  logger.log("[Gremlin Explorer] Fetching edge labels with counts...");
   const data = await gremlinFetch<RawEdgeLabelsResponse>(labelsTemplate);
 
   const values = data.result.data["@value"][0]["@value"];
@@ -187,6 +190,7 @@ const fetchEdgesAttributes = async (
   const edgesTemplate = edgesSchemaTemplate({
     types: labels,
   });
+  logger.log("[Gremlin Explorer] Fetching edges attributes...");
   const data = await gremlinFetch<RawEdgesSchemaResponse>(edgesTemplate);
 
   const edgesSchemas = data.result.data["@value"][0]["@value"];
@@ -238,6 +242,8 @@ const fetchSchema = async (
   summary?: GraphSummary
 ): Promise<SchemaResponse> => {
   if (!summary) {
+    logger.log("[Gremlin Explorer] No summary statistics");
+
     const vertices = await fetchVerticesSchema(gremlinFetch);
     const totalVertices = vertices.reduce((total, vertex) => {
       return total + (vertex.total ?? 0);
@@ -256,6 +262,8 @@ const fetchSchema = async (
     };
   }
 
+  logger.log("[Gremlin Explorer] Using summary statistics");
+
   const vertices = await fetchVerticesAttributes(
     gremlinFetch,
     summary.nodeLabels,

packages/graph-explorer/src/connector/gremlin/templates/edgesSchemaTemplate.test.ts

@@ -1,14 +1,18 @@
 import edgesSchemaTemplate from "./edgesSchemaTemplate";
+import { normalizeWithNoSpace as normalize } from "../../../utils/testing";
 
 describe("Gremlin > edgesSchemaTemplate", () => {
   it("Should return a template with the projection of each type", () => {
     const template = edgesSchemaTemplate({ types: ["route", "contain"] });
 
-    expect(template).toBe(
-      'g.E().project("route","contain")' +
-        '.by(V().bothE("route").limit(1))' +
-        '.by(V().bothE("contain").limit(1))' +
-        ".limit(1)"
+    expect(normalize(template)).toBe(
+      normalize(`
+        g.E()
+          .project("route", "contain")
+          .by(V().bothE("route").limit(1))
+          .by(V().bothE("contain").limit(1))
+          .limit(1)
+      `)
     );
   });
 });

packages/graph-explorer/src/connector/gremlin/templates/edgesSchemaTemplate.ts

@@ -1,3 +1,4 @@
+import dedent from "dedent";
 import { uniq } from "lodash";
 
 /**
@@ -14,7 +15,16 @@ import { uniq } from "lodash";
  *  .limit(1)
  */
 export default function edgesSchemaTemplate({ types }: { types: string[] }) {
-  const labels = uniq(types.flatMap(type => type.split("::")));
+  // Labels with quotes
+  const labels = uniq(types.flatMap(type => type.split("::"))).map(
+    label => `"${label}"`
+  );
 
-  return `g.E().project(${labels.map(l => `"${l}"`).join(",")})${labels.map(l => `.by(V().bothE("${l}").limit(1))`).join("")}.limit(1)`;
+
+  return dedent`
+    g.E()
+      .project(${labels.join(", ")})
+      ${labels.map(label => `.by(V().bothE(${label}).limit(1))`).join("\n      ")}
+      .limit(1)
+  `;
 }

packages/graph-explorer/src/connector/gremlin/templates/verticesSchemaTemplate.test.ts

@@ -7,10 +7,14 @@ describe("Gremlin > verticesSchemaTemplate", () => {
 
     expect(normalize(template)).toBe(
       normalize(`
-        g.V().project("airport","country")
-          .by(V().hasLabel("airport").limit(1))
-          .by(V().hasLabel("country").limit(1))
-          .limit(1)
+        g.V().union(
+          __.hasLabel("airport").limit(1),
+          __.hasLabel("country").limit(1)
+        )
+        .fold()
+        .project("airport", "country")
+        .by(unfold().hasLabel("airport"))
+        .by(unfold().hasLabel("country"))
       `)
     );
   });

packages/graph-explorer/src/connector/gremlin/templates/verticesSchemaTemplate.ts

@@ -1,20 +1,40 @@
+import dedent from "dedent";
 import { uniq } from "lodash";
 
 /**
  * Given a set of nodes labels, it returns a Gremlin template that contains
  * one sample of each node label.
  *
  * @example
- * types = ["route", "contain"]
+ * types = ["airport", "country"]
  *
  * g.V()
- *  .project("airport","country")
- *  .by(V().hasLabel("airport").limit(1))
- *  .by(V().hasLabel("country").limit(1))
- *  .limit(1)
+ *   .union(
+ *     __.hasLabel("airport").limit(1),
+ *     __.hasLabel("country").limit(1)
+ *   )
+ *   .fold()
+ *   .project(
+ *     "airport", "country"
+ *   )
+ *   .by(unfold().hasLabel("airport"))
+ *   .by(unfold().hasLabel("country"))
  */
 export default function verticesSchemaTemplate({ types }: { types: string[] }) {
-  const labels = uniq(types.flatMap(type => type.split("::")));
+  // Labels with quotes
+  const labels = uniq(types.flatMap(type => type.split("::"))).map(
+    label => `"${label}"`
+  );
 
-  return `g.V().project(${labels.map(l => `"${l}"`).join(",")})${labels.map(l => `.by(V().hasLabel("${l}").limit(1))`).join("")}.limit(1)`;
+  return dedent`
+    g.V()
+      .union(
+        ${labels.map(label => `__.hasLabel(${label}).limit(1)`).join(",\n        ")}
+      )
+      .fold()
+      .project(
+        ${labels.join(",\n        ")}
+      )
+      ${labels.map(label => `.by(unfold().hasLabel(${label}))`).join("\n      ")}
+    `;
 }

packages/graph-explorer/src/utils/index.ts

@@ -7,5 +7,6 @@ export { default as useClickOutside } from "./useClickOutside";
 export { default as sanitizeText } from "./sanitizeText";
 export { DEFAULT_SERVICE_TYPE } from "./constants";
 export { default as escapeString } from "./escapeString";
+export { default as logger } from "./logger";
 export * from "./set";
 export * from "./env";

packages/graph-explorer/src/utils/logger.ts

@@ -0,0 +1,29 @@
+/* eslint-disable no-console */
+
+import { env } from "./env";
+
+/*
+# DEV NOTE
+
+This is a simple logging utility that will allow `console.log` calls any time
+`env.DEV === true`. This will be useful for local development and debugging.
+
+I can imagine a future where this logger has some additional functionality where
+it can send errors to the server and maybe allow the use to enable debug logging
+at runtime.
+*/
+
+export default {
+  /** Calls `console.log` if the app is running in DEV mode. */
+  log(message?: any, ...optionalParams: any[]) {
+    env.DEV && console.log(message, optionalParams);
+  },
+  /** Calls `console.warn`. */
+  warn(message?: any, ...optionalParams: any[]) {
+    console.warn(message, optionalParams);
+  },
+  /** Calls `console.error`. */
+  error(message?: any, ...optionalParams: any[]) {
+    console.error(message, optionalParams);
+  },
+};