Some suggestions for changes/cleanup to the CMake build structure that would help to position better for extensions with other tooling - such as #6 - for example:

• Using a recursive arrangement and an add_directory() can make a build harder to follow for small projects, so consolidation could be considered simpler for a project with only a few source files
• The Catch2 library is built and linked, but it just needs to be included as a single header
• Explicitly setting the CMAKE_CXX_COMPILER as g++ makes the build less easy to use with clang, for example - these settings are usually best factored out to related toolchain files and referenced on the command line as -CMAKE_TOOLCHAIN_FILE=
• The file empty.cpp looks like a holdover from an earlier stage, and isn't needed now for CMake to figure out the linker language
• The bloom filter library in the CMakeLists.txt references the hpp file, which shouldn't be needed as it's not an independent translation unit

And related:

• Using Catch2 as single header usually means splitting out a separate test runner cpp file to reduce recompile times, also better for scaling up test suite sizes

The bloom filter lookups are fast (nanoseconds rather than milliseconds on ~3million records) however the binary vector load time in readVectorFromStream is slow (seconds).

Ideally, we would improve this load time while still keeping the binary small and lookup times fast. This may mean moving the underlying data structure from a bool vector which is challenging to write to file as it doesn't store its elements as a contiguous array.

Enhancement:

I appreciate in a 2-file project it's probably not so much an issue but you really shouldn't be putting using namespace std in a header file as this forcibly includes the whole std namespace for any implementation code that wishes to use the bloom.hpp header.

Avoiding placing using namespace std (or indeed any using namespace X) in a header is mentioned in so many places I forget where - maybe Exceptional C++ - / Sutter would be one, Scott Meyers' Effective STL almost certainly another (Bjarne Stroustrup C++ Prog Lang another?) - but one ref I did just find an old classic from Sutter's GOTW: http://gotw.ca/gotw/053.htm see Section 2, Guideline #1:

1. Using directives should be avoided entirely, especially in header files

Sutter also makes the distinction here - as I did above - in a footnote that this guideline is valid for shared header files (i.e. used by more than one cpp) : http://gotw.ca/gotw/053.htm#1 so you could argue "no change here" since only one cpp, but then you could also argue that it is simply poor style - which I why I label this as an enhancement

BTW: nothing against using namespace std in your cpp, should you wish to do so - so this is probably a one-line change: remove using namespace std from the header and add it to the cpp. Depending on where you place the using in your cpp, you may need to fully qualify any STL types in the header file.

I can fork this repo and supply a PR with change if you accept this as an (enhancement) issue.