Merge pull request #139 from effigies/feat/tsv-maxrows

feat: Process TSV files as streams and validate only the first 1000 rows by default

deno.json

@@ -41,6 +41,7 @@
     "@std/io": "jsr:@std/[email protected]",
     "@std/log": "jsr:@std/[email protected]",
     "@std/path": "jsr:@std/[email protected]",
+    "@std/streams": "jsr:@std/[email protected]",
     "@std/yaml": "jsr:@std/yaml@^1.0.4"
   },
   "tasks": {

src/files/deno.test.ts

@@ -3,7 +3,8 @@ import { readAll, readerFromStreamReader } from '@std/io'
 import { basename, dirname, fromFileUrl, join } from '@std/path'
 import { EOL } from '@std/fs'
 import type { FileTree } from '../types/filetree.ts'
-import { BIDSFileDeno, readFileTree, UnicodeDecodeError } from './deno.ts'
+import { BIDSFileDeno, readFileTree } from './deno.ts'
+import { UnicodeDecodeError } from './streams.ts'
 import { requestReadPermission } from '../setup/requestPermissions.ts'
 import { FileIgnoreRules } from './ignore.ts'
 

src/files/deno.ts

@@ -7,18 +7,9 @@ import { type BIDSFile, FileTree } from '../types/filetree.ts'
 import { requestReadPermission } from '../setup/requestPermissions.ts'
 import { FileIgnoreRules, readBidsIgnore } from './ignore.ts'
 import { logger } from '../utils/logger.ts'
+import { createUTF8Stream } from './streams.ts'
 export { type BIDSFile, FileTree }
 
-/**
- * Thrown when a text file is decoded as UTF-8 but contains UTF-16 characters
- */
-export class UnicodeDecodeError extends Error {
-  constructor(message: string) {
-    super(message)
-    this.name = 'UnicodeDecode'
-  }
-}
-
 /**
  * Deno implementation of BIDSFile
  */
@@ -67,27 +58,17 @@ export class BIDSFileDeno implements BIDSFile {
    * Read the entire file and decode as utf-8 text
    */
   async text(): Promise<string> {
-    const streamReader = this.stream
-      .pipeThrough(new TextDecoderStream('utf-8'))
-      .getReader()
-    let data = ''
+    const reader = this.stream.pipeThrough(createUTF8Stream()).getReader()
+    const chunks: string[] = []
     try {
-      // Read once to check for unicode issues
-      const { done, value } = await streamReader.read()
-      // Check for UTF-16 BOM
-      if (value && value.startsWith('\uFFFD')) {
-        throw new UnicodeDecodeError('This file appears to be UTF-16')
-      }
-      if (done) return data
-      data += value
-      // Continue reading the rest of the file if no unicode issues were found
       while (true) {
-        const { done, value } = await streamReader.read()
-        if (done) return data
-        data += value
+        const { done, value } = await reader.read()
+        if (done) break
+        chunks.push(value)
       }
+      return chunks.join('')
     } finally {
-      streamReader.releaseLock()
+      reader.releaseLock()
     }
   }
 

src/files/filetree.ts

@@ -5,7 +5,11 @@ import { FileIgnoreRules } from './ignore.ts'
 
 const nullFile = {
   size: 0,
-  stream: new ReadableStream(),
+  stream: new ReadableStream({
+    start(controller) {
+      controller.close()
+    }
+  }),
   text: () => Promise.resolve(''),
   readBytes: async (size: number, offset?: number) => new Uint8Array(),
   parent: new FileTree('', '/'),

src/files/json.test.ts

@@ -1,5 +1,4 @@
 import { type assert, assertObjectMatch } from '@std/assert'
-import type { BIDSFileDeno, UnicodeDecodeError } from './deno.ts'
 import type { BIDSFile } from '../types/filetree.ts'
 import type { FileIgnoreRules } from './ignore.ts'
 

src/files/streams.test.ts

@@ -0,0 +1,37 @@
+import { assert, assertEquals } from '@std/assert'
+import { createUTF8Stream, UnicodeDecodeError } from './streams.ts'
+import { streamFromUint8Array, streamFromString } from '../tests/utils.ts'
+
+Deno.test('createUTF8Stream', async (t) => {
+  await t.step('should return a TransformStream with UTF8StreamTransformer', () => {
+    const stream = createUTF8Stream()
+    assertEquals(stream instanceof TransformStream, true)
+  })
+
+  await t.step('should correctly transform UTF-8 input', async () => {
+    const rawstream = streamFromString('Hello, world!')
+    const reader = rawstream.pipeThrough(createUTF8Stream()).getReader()
+    const { value } = await reader.read()
+    assertEquals(value, 'Hello, world!')
+
+    await reader.cancel()
+  })
+
+  await t.step('should throw UnicodeDecodeError for UTF-16 input', async () => {
+    const rawStream = streamFromUint8Array(new Uint8Array([0xFF, 0xFE, 0x00, 0x00]))
+
+    let reader
+    try {
+      // The exception can't be localized to either of the following lines
+      // but is raised before the second returns
+      reader = rawStream.pipeThrough(createUTF8Stream()).getReader()
+      const { value } = await reader.read()
+      assert(false, 'Expected UnicodeDecodeError, got ' + value)
+    } catch (e: any) {
+      assertEquals(e instanceof UnicodeDecodeError, true)
+      assertEquals(e?.message, 'This file appears to be UTF-16')
+    } finally {
+      if (reader) await reader.cancel
+    }
+  })
+})

src/files/streams.ts

@@ -0,0 +1,51 @@
+/**
+ * Thrown when a text file is decoded as UTF-8 but contains UTF-16 characters
+ */
+export class UnicodeDecodeError extends Error {
+  constructor(message: string) {
+    super(message)
+    this.name = 'UnicodeDecode'
+  }
+}
+
+/**
+ * A transformer that ensures the input stream is valid UTF-8 and throws
+ * a UnicodeDecodeError if UTF-16 BOM is detected
+ */
+export class UTF8StreamTransformer implements Transformer<Uint8Array, string> {
+  private decoder: TextDecoder
+  private firstChunk: boolean
+
+  constructor() {
+    this.decoder = new TextDecoder('utf-8')
+    this.firstChunk = true
+  }
+
+  transform(chunk: Uint8Array, controller: TransformStreamDefaultController<string>) {
+    // Check first chunk for UTF-16 BOM
+    if (this.firstChunk) {
+      const decoded = this.decoder.decode(chunk, { stream: true })
+      if (decoded.startsWith('\uFFFD')) {
+        throw new UnicodeDecodeError('This file appears to be UTF-16')
+      }
+      this.firstChunk = false
+      controller.enqueue(decoded)
+    } else {
+      controller.enqueue(this.decoder.decode(chunk, { stream: true }))
+    }
+  }
+
+  flush(controller: TransformStreamDefaultController<string>) {
+    const final = this.decoder.decode()
+    if (final) {
+      controller.enqueue(final)
+    }
+  }
+}
+
+/**
+ * Creates a TransformStream that validates and decodes UTF-8 text
+ */
+export function createUTF8Stream() {
+  return new TransformStream(new UTF8StreamTransformer())
+}