B Programming Language: History, Features, and Modern Reconstruction

B emerged as a simplified version of BCPL, often described as “BCPL filtered through Ken Thompson’s brain.” The language was created to run on the PDP‑7, a small machine with severe memory constraints. There was never an “A” language; the name B derives directly from its predecessor, BCPL.

Technical Characteristics

B operates with a single data type: the machine word. It lacks explicit types such as int or char and does not provide pointers. Because there is no native byte type, extracting a character from a string requires a dedicated char function.

The language uses threaded code, a technique similar to Forth, to keep the compiler footprint tiny. The auto keyword, now familiar from C++, originated in B to declare variables on the stack. Early versions of B omitted a for loop, so programmers relied on while loops for iteration.

Reconstruction Process

The original B compiler disappeared, prompting a reverse‑engineering effort to make B code runnable again. By analyzing surviving object code, standard library functions such as printf, and existing documentation, a new compiler was built from scratch. The reconstructed compiler consists of roughly 1,000 lines of B code and is capable of compiling itself.

Because the original intermediate code format was forgotten, an invented intermediate format bridges the source and the B assembler. The assembler then produces output compatible with the Unix assembler.

Mechanisms & Explanations

Compilation Process – B source is fed to the B compiler, which emits intermediate code. This code passes to a B assembler, generating object files that the Unix assembler can link.

Threaded Code – The compiler emits threaded code for a stack machine. This design choice stems from the original compiler’s need to fit into a very limited memory space, precluding complex optimizations.

Portability – Relying solely on machine words and a threaded code model makes B highly portable. Modern implementations have run on PDP‑11, PDP‑10, RISC‑V, MIPS, and x86 architectures.

Portability in Practice

The language’s minimalism—single data type, lack of pointers, and threaded code—allows the same B source to be compiled across diverse hardware. This portability aligns with the original design goal of running on constrained machines while still being adaptable to contemporary platforms.

Notable Quotations

• “B was actually uh probably called B because it was a simplified version of BCPL.”
• “B was like BCPL filtered through Ken Thompson’s brain.”
• “One interesting feature of B is that it only has a single data type just the machine word.”
• “The compiler is very short uh because it had to fit into a small memory uh amount of memory.”
• “You can write a compiler in like about thousand lines of B code that compiles itself.”