Our C++ code repository is available at: https://github.com/abseil/abseil-cpp
By adopting these new Apache-licensed libraries, you can reap the benefit of years (over a decade in many cases) of our design and optimization work in this space. Our past experience is baked in.
Just as interesting, we’ve also prepared for the future: several types in Abseil’s C++ libraries are “pre-adoption” versions of C++17 types like string_view and optional - implemented in C++11 to the greatest extent possible. We look forward to moving more and more of our code to match the current standard, and using these new vocabulary types helps us make that transition. Importantly, in C++17 mode these types are merely aliases to the standard, ensuring that you only ever have one type for optional or string_view in a project at a time. Put another way: Abseil is focused on the engineering task of providing APIs that remain stable over time.
Consisting of the foundational C++ and Python code at Google, Abseil includes libraries that will grow to underpin other Google-backed open source projects like gRPC, Protobuf and TensorFlow. We love those projects, and we love the users of those projects - we want to ensure smooth usage for these things over time. In the next few months we’ll introduce new distribution methods to incorporate these projects as a collection into your project.
Continuing with the “over time” theme, Abseil aims for compatibility with major compilers, platforms and standard libraries for approximately 5 years. Our 5-year target also applies to language version: we assume everyone builds with C++11 at this point. (In 2019 we’ll start talking about requiring C++14 as our base language version.) This 5-year horizon is part of our balance between “support everything” and “provide modern implementations and APIs.”
Highlights of the initial release include:
- Zero configuration: most platforms (OS, compiler, architecture) should just work.
- Pre-adoption for C++17 types: string_view, optional, any. We’ll follow up with variant soon.
- Our primary synchronization type, absl::Mutex, has an elegant interface and has been extensively optimized.
- Efficient support for handling time: absl::Time and absl::Duration are conceptually similar to std::chrono types, but are concrete (not class templates) and have defined behavior in all cases. Additionally, our clock-sampling API absl::Now() is more heavily optimized than most standard library calls for std::chrono::system_clock::now().
- String handling routines: among internal users, we’ve been told that releasing absl::StrCat(), absl::StrJoin(), and absl::StrSplit() would itself be a big improvement for the open source C++ world.
By Titus Winters, Abseil Lead
