infix-calculator

A typical command-line infix calculator.

Intended as a hands-on C++ project to get familiar with Flex and Bison.

Features

Can process semicolon-terminated expression input from both stdin and text files. Supports boolean, integral, and floating point expressions using standard arithmetic and logical operations, with C-style operator precedence. Allows line comments within input files and grouping of subexpressions with parentheses. Variable assignment is supported and some builtin functions are provided, e.g. logarithmic, exponential, trigonometric, binary min + max.

The calculator parser is built as a separate static or shared library that the calculator command-line tool links against. PIMPL is used to provide a stable ABI and prevent leaking Flex + Bison generated names and types into user code.

Below is a sample program that demonstrates some of the available features:

a = 1.5;
b = sin(1.5) * cos(1.8) + log(a);
b += exp(a - b);
# non-assignment expressions write result to output stream
sqrt(b);

Running the command-line tool on this input should print the following:

<double> 1.98157

Building + Installing

Requires CMake >=3.16, Flex >=2.6.2, and Bison >=3.5.1. Flex and Bison can either be installed using your system package manager, e.g. for *nix systems, with an alternative being setting either environment or CMake variables with the directory location of Flex and Bison. On Windows, it is recommended to use WinFlexBison, a port of Flex and Bison to Windows, as the original tools were designed only for distribution for GNU/Linux style systems.

*nix

Building is easy with the provided build.sh build script. For usage, type

./build.sh --help

To build release binaries for this project, simply use the command

./build.sh -c Release

Simply typing ./build.sh will build unoptimized binaries with debug symbols.

To install to /usr/local, your prefix of choice, use the command

cmake --install build

For a custom installation root, use the --prefix argument, e.g.

cmake --install build --prefix /path/to/install/location

Windows

Building is easy with the provided build.bat build script. For usage, type

build --help

To build release binaries for this project, simple use the command

build -c Release

Simply typing build will build unoptimized binaries and the program database with debugging info. You can specify the target architecture using the -a flag, e.g. to build 32-bit release binaries instead of the default 64-bit ones, use

build -a x86 -c Release

Currently, the Visual Studio toolset used will be whichever is the default.

To install to a custom installation root, use the --prefix argument, e.g.

cmake --install build_windows_x64 --prefix %USERPROFILE%\pdcalc-master

Usage from CMake

After build + installation, the calculator parser library can be used from CMake in the normal fashion with CMake's find_package command. To locate the library, use the following in a CMakeLists.txt or CMake script:

find_package(pdcalc 0.1.0)

If the library is found, one can write programs using the headers, for example:

/**
 * @file pdcalc_ex.cc
 * @author Derek Huang
 * @brief C++ minimal example using pdcalc
 * @copyright MIT License
 */

#include <filesystem>

#include <pdcalc/calc_parser.hh>  // note: no main API header yet

int main()
{
  // read input from stdin and write to stdout (returns true on success)
  pdcalc::calc_parser parser;
  return !parser();
}

Then, when adding the CMake target for this program, one can to specify the compile + link requirements simply via use of target_link_libraries without needing to manually update the include or link directories for the project or target. Using the above trivial program as an example, all one needs to do in a CMakeLists.txt is the following:

add_executable(pdcalc_ex pdcalc_ex.cc)
target_link_libraries(pdcalc_ex PRIVATE pdcalc::pdcalc)

On a *nix system, assuming the build directory is build, one can run the produced example program as follows:

echo "a = sin(1.9); tan(exp(-a));" | ./build/pdcalc_ex

The following should then be written to standard output:

<double> 0.408923