fix: data-to-be-packed resize remove existing values (#30)

Context:
- When the number of exception values `bestbbestcexceptmaxb[1]` +
`data_pointers[index as usize]` exceeded the pre-allocated size of
`data_to_be_packed`, the resize logic replaced the entire buffer,
discarding existing values.

Implemented:
- Fixed the resizing logic to preserve existing exception values by
extending the vector with zeros instead of overwriting.
- Added `test_exception_value_vector_resizes` unit test.

Co-authored-by: Weixing Zhang <[email protected]>

Author: Weixing-Zhang

src/rust/integer_compression/fastpfor.rs

@@ -168,9 +168,7 @@ impl FastPFOR {
                             + self.bestbbestcexceptmaxb[1] as usize)
                             as u32;
                     new_size = helpers::greatest_multiple(new_size + 31, 32);
-                    self.data_to_be_packed[index as usize].resize(new_size as usize, 1);
-                    self.data_to_be_packed[index as usize]
-                        .clone_from_slice(&vec![1; new_size as usize]);
+                    self.data_to_be_packed[index as usize].resize(new_size as usize, 0);
                 }
                 for k in 0..self.block_size {
                     if (input[(k + tmp_input_offset) as usize] >> self.bestbbestcexceptmaxb[0]) != 0

tests/basic_tests.rs

@@ -415,3 +415,45 @@ fn test_fastpfor_headless_compress_unfit_pagesize() {
 
     assert_eq!(input, decoded, "Input and decompressed data do not match");
 }
+
+#[test]
+fn test_exception_value_vector_resizes() {
+    let page_size = 512;
+    let test_input_size = page_size * 2;
+
+    // every even index value is large which will trigger exception buffer to be resize
+    let input: Vec<u32> = (0..test_input_size)
+        .map(|i| if i % 2 == 0 { 1 << 30 } else { 3 })
+        .collect();
+
+    let mut output: Vec<u32> = vec![0; input.len() * 4];
+    let mut decoded: Vec<u32> = vec![0; input.len()];
+    let mut input_offset = Cursor::new(0u32);
+    let mut output_offset = Cursor::new(0u32);
+
+    let mut codec = FastPFOR::new(page_size, BLOCK_SIZE_128);
+    codec
+        .compress(
+            &input,
+            input.len() as u32,
+            &mut input_offset,
+            &mut output,
+            &mut output_offset,
+        )
+        .expect("compression failed");
+
+    let compressed_len = output_offset.position() as usize;
+    input_offset.set_position(0);
+
+    codec
+        .uncompress(
+            &output,
+            compressed_len as u32,
+            &mut input_offset,
+            &mut decoded,
+            &mut Cursor::new(0u32),
+        )
+        .expect("decompression failed");
+
+    assert_eq!(input, decoded, "Input and decompressed data do not match");
+}