fix: Zip CSV uploads not working (#1457)

Fixes #1080. Fixes #1416 

Updates jszip and uses the internal stream helper instead of nodestream
which we would need a polyfill for. The progress on zip uploads is a bit
odd since we use the unzip progress. It seems papaparse loads more from
the stream while processing/before processing and can finish reading the
zip before it's done processing chunks.

Also, uploading the tables seems to be a blocking operation. Not sure if
that's an easy fix, but after the 50% mark on uploading a large zip,
there are some blocks on the main thread (marching ants freezes as an
indicator).

There is a ~2GB limit for JSZip it seems
Stuk/jszip#777

package.json

@@ -82,7 +82,6 @@
     "@types/eslint": "^8.4.10",
     "@types/jest": "^29.2.5",
     "@types/jquery": "^3.5.14",
-    "@types/jszip": "3.1.7",
     "@types/lodash": "^4.14.182",
     "@types/lodash.clamp": "^4.0.6",
     "@types/lodash.debounce": "^4.0.6",

packages/code-studio/package.json

@@ -39,7 +39,7 @@
     "@fortawesome/react-fontawesome": "^0.2.0",
     "classnames": "^2.3.1",
     "event-target-shim": "^6.0.2",
-    "jszip": "3.2.2",
+    "jszip": "3.10.1",
     "lodash.debounce": "^4.0.8",
     "lodash.throttle": "^4.1.1",
     "memoize-one": "^5.1.1",
@@ -58,10 +58,6 @@
     "redux-thunk": "^2.4.1",
     "shortid": "^2.2.16"
   },
-  "dependenciesComments": {
-    "@types/jszip": "3.1.7 is the closest to 3.2.2 available. JSZip adds official typings in 3.4.0, so remove if going past JSZip 3.4.0",
-    "jszip": "Pinned to 3.2.2 b/c 3.3.0+ breaks nodestream usage. Not fixed as of 3.5.0. https://github.com/Stuk/jszip/issues/663"
-  },
   "homepage": ".",
   "main": "build/index.html",
   "files": [

packages/console/src/csv/CsvInputBar.tsx

@@ -178,7 +178,7 @@ class CsvInputBar extends Component<CsvInputBarProps, CsvInputBarState> {
     }
   }
 
-  handleFile(file: Blob | JSZipObject, isZip = false): void {
+  handleFile(file: File | Blob | JSZipObject, isZip = false): void {
     log.info(
       `Starting CSV parser for ${
         file instanceof File ? file.name : 'pasted values'

packages/console/src/csv/CsvParser.ts

@@ -5,6 +5,7 @@ import type { IdeSession, Table } from '@deephaven/jsapi-types';
 import type { JSZipObject } from 'jszip';
 import CsvTypeParser from './CsvTypeParser';
 import { CsvTypes } from './CsvFormats';
+import makeZipStreamHelper from './ZipStreamHelper';
 
 const log = Log.module('CsvParser');
 
@@ -15,7 +16,7 @@ const ZIP_CONSOLIDATE_CHUNKS = 650;
 interface CsvParserConstructor {
   onFileCompleted: (tables: Table[]) => void;
   session: IdeSession;
-  file: Blob | JSZipObject;
+  file: File | Blob | JSZipObject;
   type: CsvTypes;
   readHeaders: boolean;
   onProgress: (progressValue: number) => boolean;
@@ -70,6 +71,8 @@ class CsvParser {
     this.onProgress = onProgress;
     this.onError = onError;
     this.tables = [];
+    this.rowCount = 0;
+    this.rowsProcessed = 0;
     this.chunks = 0;
     this.totalChunks = isZip
       ? 0
@@ -102,7 +105,7 @@ class CsvParser {
 
   session: IdeSession;
 
-  file: Blob | JSZipObject;
+  file: File | Blob | JSZipObject;
 
   isZip: boolean;
 
@@ -122,6 +125,10 @@ class CsvParser {
 
   types?: string[];
 
+  rowCount: number;
+
+  rowsProcessed: number;
+
   chunks: number;
 
   totalChunks: number;
@@ -167,17 +174,20 @@ class CsvParser {
   }
 
   parse(): void {
-    const handleParseDone = (types: string[]) => {
-      const toParse = this.isZip
-        ? (this.file as JSZipObject).nodeStream(
-            // JsZip types are incorrect, thus the funny casting
-            // Actual parameter is 'nodebuffer'
-            'nodebuffer' as 'nodestream',
-            this.handleNodeUpdate
-          )
-        : (this.file as Blob);
+    const handleParseDone = (types: string[], rowCount: number) => {
       this.types = types;
-      Papa.parse(toParse, this.config);
+      this.rowCount = rowCount;
+
+      if (this.file instanceof File || this.file instanceof Blob) {
+        Papa.parse(this.file, this.config);
+      } else {
+        const zipStream = makeZipStreamHelper(this.file, this.handleNodeUpdate);
+        // This is actually a stream, but papaparse TS doesn't like it
+        Papa.parse(zipStream as unknown as Blob, this.config);
+        // The stream needs to be manually resumed since jszip starts paused
+        // Papaparse does not call resume and assumes the stream is already reading
+        zipStream.resume();
+      }
     };
     const typeParser = new CsvTypeParser(
       handleParseDone,
@@ -274,6 +284,9 @@ class CsvParser {
     }
     assertNotNull(this.headers);
     assertNotNull(types);
+
+    this.rowsProcessed += columns[0].length;
+
     session
       .newTable(this.headers, types, columns, this.timeZone)
       .then(table => {
@@ -294,7 +307,9 @@ class CsvParser {
         if (totalChunks > 0) {
           progress = Math.round((tables.length / totalChunks) * 50) + 50;
         } else {
-          progress = Math.round(50 + this.zipProgress / 2);
+          // The zip file can be read entirely while in the middle of parsing
+          // Since we know the number of rows from the type parsing, use that for progress
+          progress = Math.round((this.rowsProcessed / this.rowCount) * 50) + 50;
         }
         log.debug2(`CSV parser progress ${progress}`);
         onProgress(progress);

packages/console/src/csv/CsvTypeParser.ts

@@ -4,6 +4,7 @@ import Papa, { Parser, ParseResult, ParseLocalConfig } from 'papaparse';
 // Intentionally using isNaN rather than Number.isNaN
 /* eslint-disable no-restricted-globals */
 import NewTableColumnTypes from './NewTableColumnTypes';
+import makeZipStreamHelper from './ZipStreamHelper';
 
 // Initially column types start as unknown
 const UNKNOWN = 'unknown';
@@ -156,8 +157,8 @@ class CsvTypeParser {
   }
 
   constructor(
-    onFileCompleted: (types: string[]) => void,
-    file: Blob | JSZipObject,
+    onFileCompleted: (types: string[], rowCount: number) => void,
+    file: File | Blob | JSZipObject,
     readHeaders: boolean,
     parentConfig: ParseLocalConfig<unknown, Blob | NodeJS.ReadableStream>,
     nullString: string | null,
@@ -175,6 +176,7 @@ class CsvTypeParser {
     this.onError = onError;
     this.chunks = 0;
     this.totalChunks = totalChunks;
+    this.rowCount = 0;
     this.isZip = isZip;
     this.shouldTrim = shouldTrim;
     this.zipProgress = 0;
@@ -192,9 +194,9 @@ class CsvTypeParser {
     };
   }
 
-  onFileCompleted: (types: string[]) => void;
+  onFileCompleted: (types: string[], rowCount: number) => void;
 
-  file: Blob | JSZipObject;
+  file: File | Blob | JSZipObject;
 
   readHeaders: boolean;
 
@@ -210,6 +212,8 @@ class CsvTypeParser {
 
   totalChunks: number;
 
+  rowCount: number;
+
   isZip: boolean;
 
   shouldTrim: boolean;
@@ -219,15 +223,16 @@ class CsvTypeParser {
   config: ParseLocalConfig<unknown, Blob | NodeJS.ReadableStream>;
 
   parse(): void {
-    const toParse = this.isZip
-      ? (this.file as JSZipObject).nodeStream(
-          // JsZip types are incorrect, thus the funny casting
-          // Actual parameter is 'nodebuffer'
-          'nodebuffer' as 'nodestream',
-          this.handleNodeUpdate
-        )
-      : (this.file as Blob);
-    Papa.parse(toParse, this.config);
+    if (this.file instanceof File || this.file instanceof Blob) {
+      Papa.parse(this.file, this.config);
+    } else {
+      const zipStream = makeZipStreamHelper(this.file, this.handleNodeUpdate);
+      // This is actually a stream, but papaparse TS doesn't like it
+      Papa.parse(zipStream as unknown as Blob, this.config);
+      // The stream needs to be manually resumed since jszip starts paused
+      // Papaparse does not call resume and assumes the stream is already reading
+      zipStream.resume();
+    }
   }
 
   handleChunk(result: ParseResult<string[]>, parser: Parser): void {
@@ -245,6 +250,8 @@ class CsvTypeParser {
       }
     }
 
+    this.rowCount += data.length;
+
     assertNotNull(this.types);
 
     const cloneTypes = [...this.types];
@@ -294,7 +301,8 @@ class CsvTypeParser {
           type === UNKNOWN || type === NewTableColumnTypes.LOCAL_TIME
             ? NewTableColumnTypes.STRING
             : type
-        )
+        ),
+        this.rowCount
       );
     }
   }

packages/console/src/csv/ZipStreamHelper.ts

@@ -0,0 +1,33 @@
+import type { JSZipObject, OnUpdateCallback, JSZipStreamHelper } from 'jszip';
+
+/**
+ * This is used to help papaparse understand our stream.
+ * It uses these fields for feature detection, but never actually calls read()
+ * https://github.com/mholt/PapaParse/blob/master/papaparse.js#L244
+ */
+interface ZipStreamHelper extends JSZipStreamHelper<string> {
+  readable: boolean;
+  read(): void;
+  removeListener(): void;
+}
+
+export default function makeZipStreamHelper(
+  zipObj: JSZipObject,
+  onUpdate: OnUpdateCallback
+) {
+  const helper: ZipStreamHelper = (
+    zipObj as JSZipObject & {
+      // The type could be anything except nodebuffer from https://stuk.github.io/jszip/documentation/api_zipobject/internal_stream.html
+      // We only need it as a string though
+      // JSZip types don't include this method for some reason
+      internalStream(type: 'string'): JSZipStreamHelper<string>;
+    }
+  ).internalStream('string') as ZipStreamHelper;
+
+  helper.readable = true;
+  helper.read = () => false;
+  helper.removeListener = () => false;
+  helper.on('data', (_, metadata) => onUpdate(metadata));
+
+  return helper;
+}