Merge pull request #127 from effigies/fix/nifti_headers

fix: Truncate NIfTI header data to avoid extensions

changelog.d/20241204_103403_effigies_nifti_headers.md

@@ -0,0 +1,49 @@
+<!--
+A new scriv changelog fragment.
+
+Uncomment the section that is right (remove the HTML comment wrapper).
+-->
+
+<!--
+### Added
+
+- A bullet item for the Added category.
+
+-->
+<!--
+### Changed
+
+- A bullet item for the Changed category.
+
+-->
+### Fixed
+
+- Resolve issue with parsing headers of NIfTI files with large extensions.
+  Fixes [issue 126].
+
+[issue 126]: https://github.com/bids-standard/bids-validator/issues/126
+
+<!--
+### Deprecated
+
+- A bullet item for the Deprecated category.
+
+-->
+<!--
+### Removed
+
+- A bullet item for the Removed category.
+
+-->
+<!--
+### Security
+
+- A bullet item for the Security category.
+
+-->
+<!--
+### Infrastructure
+
+- A bullet item for the Infrastructure category.
+
+-->

src/files/deno.ts

@@ -93,14 +93,17 @@ export class BIDSFileDeno implements BIDSFile {
 
   /**
    * Read bytes in a range efficiently from a given file
+   *
+   * Reads up to size bytes, starting at offset.
+   * If EOF is encountered, the resulting array may be smaller.
    */
   async readBytes(size: number, offset = 0): Promise<Uint8Array> {
     const handle = this.#openHandle()
     const buf = new Uint8Array(size)
     await handle.seek(offset, Deno.SeekMode.Start)
-    await handle.read(buf)
+    const nbytes = await handle.read(buf) ?? 0
     handle.close()
-    return buf
+    return buf.subarray(0, nbytes)
   }
 
   /**

src/files/nifti.test.ts

@@ -53,4 +53,23 @@ Deno.test('Test loading nifti header', async (t) => {
     })
     assertObjectMatch(error, { key: 'NIFTI_HEADER_UNREADABLE' })
   })
+
+  await t.step('Tolerate big headers', async () => {
+    const path = 'big_header.nii.gz'
+    const root = './tests/data/'
+    const file = new BIDSFileDeno(root, path, ignore)
+    let error: any = undefined
+    const header = await loadHeader(file)
+    assert(header !== undefined)
+    assertObjectMatch(header, {
+      dim: [3, 1, 1, 1, 1, 1, 1],
+      pixdim: [1, 1, 1, 1, 1, 1, 1],
+      shape: [1, 1, 1],
+      voxel_sizes: [1, 1, 1],
+      dim_info: { freq: 0, phase: 0, slice: 0 },
+      xyzt_units: { xyz: 'unknown', t: 'unknown' },
+      qform_code: 0,
+      sform_code: 2,
+    })
+  })
 })

src/files/nifti.ts

@@ -1,4 +1,4 @@
-import { isCompressed, readHeader } from '@mango/nifti'
+import { isCompressed, isNIFTI1, isNIFTI2, NIFTI1, NIFTI2 } from '@mango/nifti'
 import type { BIDSFile } from '../types/filetree.ts'
 import { logger } from '../utils/logger.ts'
 import type { NiftiHeader } from '@bids/schema/context'
@@ -11,27 +11,39 @@ async function extract(buffer: Uint8Array, nbytes: number): Promise<Uint8Array>
   const stream = new ReadableStream({
     start(controller) {
       controller.enqueue(buffer)
+      controller.close()
     },
   })
   const reader = stream.pipeThrough(new DecompressionStream('gzip')).getReader()
   let offset = 0
-  while (offset < nbytes) {
-    const { value, done } = await reader.read()
-    if (done) {
-      break
+  try {
+    while (offset < nbytes) {
+      const { value, done } = await reader.read()
+      if (done || !value) {
+        break
+      }
+      result.set(value.subarray(0, Math.min(value.length, nbytes - offset)), offset)
+      offset += value.length
     }
-    result.set(value.subarray(0, Math.min(value.length, nbytes - offset)), offset)
-    offset += value.length
+  } finally {
+    await reader.cancel()
   }
-  await reader.cancel()
-  return result
+  return result.subarray(0, offset)
 }
 
 export async function loadHeader(file: BIDSFile): Promise<NiftiHeader> {
   try {
     const buf = await file.readBytes(1024)
-    const data = isCompressed(buf.buffer) ? await extract(buf, 540) : buf
-    const header = readHeader(data.buffer)
+    const data = isCompressed(buf.buffer) ? await extract(buf, 540) : buf.slice(0, 540)
+    let header
+    if (isNIFTI1(data.buffer)) {
+      header = new NIFTI1()
+      // Truncate to 348 bytes to avoid attempting to parse extensions
+      header.readHeader(data.buffer.slice(0, 348))
+    } else if (isNIFTI2(data.buffer)) {
+      header = new NIFTI2()
+      header.readHeader(data.buffer)
+    }
     if (!header) {
       throw { key: 'NIFTI_HEADER_UNREADABLE' }
     }