Bazel Overview

Bazel is a build tool which coordinates builds and runs tests. The extension language allows it to work with source files written in any language, with native support for Java, C, C++ and Python. Bazel produces builds and runs tests for multiple platforms.

BUILD files use a simple declarative language

Bazel’s BUILD files describe how Bazel should build your project. They have a declarative structure and use a language similar to Python. BUILD files allow you to work at a high level of the system by listing rules and their attributes. The complexity of the build process is handled by these pre-existing rules. You can modify rules to tweak the build process, or write new rules to extend Bazel to work with any language or platform.

Below is the content of one of the BUILD files from a Hello World program. The two rules used here are cc_library and cc_binary.

    name = "hello-time",
    srcs = [""],
    hdrs = ["hello-time.h"],

    name = "hello-world",
    srcs = [""],
    deps = [

The dependency graph describes the entire system

Build dependencies are declared explicitly in the BUILD files, allowing Bazel to create an accurate dependency graph of the entire source code. The graph is maintained in memory, and incremental builds and parallel execution are possible because of this accurate dependency graph.

Here’s the graph of the target ‘hello-world’ from the BUILD file above:

Dependency graph of a hello-world target

Bazel’s query language allows you to produce images of the graph like the one above. You can also use the query language to access information about build dependencies and their relationships.

Build and tests are fast, correct, and reproducible

Hermetic rules and sandboxing allow Bazel to produce correct, reproducible artifacts and test results. Caching allows reuse of build artifacts and test results.

Bazel’s builds are fast. Incremental builds allow Bazel to do the minimum required work for a rebuild or retest. Correct and reproducible builds allow Bazel to reuse cached artifacts for whatever is not changed. If you change a library, Bazel will not rebuild your entire source.

Confidence in these correct results also means that you will never need to run bazel clean. If you ever need to run bazel clean, there’s a bug in Bazel.