Why BSON is not using uint32_t instead of int32_t for document lenght

Viewed 162

According to BSON specification an int32_t is used for the total number of bytes comprising the document.

I can also see that mongo-c-driver is using Libbson and it's using int32_t.

But what is the reason to use signed integer instead of unsigned integer?

the document size will never be below 0. Can someone please explain the reason?

1 Answers

Document size is limited to 16 MB, which fits in a signed 32-bit integer with room to spare.

While document size limits larger than 16 MB were considered, my guess is there was no serious consideration of document sizes exceeding 2 GB, therefore the difference in range afforded by using unsigned 32-bit integers for length wasn't a meaningful benefit.

Thus, there is no reason to use an unsigned type.

Reasons to NOT use an unsigned type:

  • Wraparound potential from UINT_MAX to 0.
  • -1 could be used as a special designator.
Related