@jeanlauliac/upd

upd

👟 Update files, fast.

Description

upd is a command-line utility that automatically determines what files in a directory need to be updated as resulting from the transformation of other files, and issues the commands to update them. In that regard it is similar to the make utility. upd focuses on convenience and performance. It can automatically identify sets of source files based on globbing patterns. It can automatically track transitive dependencies of a file to be updated.

For example, upd can be used to manage conveniently the compilation of a C++ project. I will run the process to compile all the *.cpp source files of a specified directory, then run the process to link the object files. It will track the header dependencies of each source files automatically, so that changes to a header file triggers a new compilation. At the same time, unrelated header or source changes will not cause a new compilation, making the update as fast as possible. Finally, upd take care of creating output directories before they get generated.

Get Started

The most convenient way to install upd is by using either npm or yarn. The utility is split in two main packages. upd-cli is the global CLI tool, while upd can be installed locally and separately for each project. This allows you to use completely different versions of upd on different projects easily. This constrasts with utilities such as make, which are generally global.

For installing with npm:

npm install -g @jeanlauliac/upd-cli    # only if not installed already on your machine
npm install @jeanlauliac/upd --save-dev

Then you can initialize your project by running the following in the project's root directory:

upd init

This creates a single file .updroot that is used by upd to know what is the root directory of the project. Only files in that directory or in any of its children directories can be managed and updated by upd for that particular project.

Manifest

To know which files to update and how, upd needs a document describing the relationship between files. For example, it can describe that all *.cpp files should be compiled into object files. This document is called the manifest, and should be contained in a file at the root directory of your project, under the name updfile.json. The manifest is in the JSON format, that is fairly readable by humans but tedious to maintain by hand. On the other hand it is easy to generate from a script. As such it is recommended, but not required, to use upd-configure to generate the manifest.

Here's is an example of manifest that compiles all files matching src/*.cpp into their respective object files in the output directory.

{
  "command_line_templates": [
    {
      "binary_path": "clang++",
      "arguments": [
        {
          "literals": ["-c", "-o"],
          "variables": ["output_file"]
        },
        {
          "literals": ["-I", "/usr/local/include"],
          "variables": ["input_files"]
        }
      ]
    }
  ],
  "source_patterns": ["src/(*).cpp"],
  "rules": [
    {
      "command_line_ix": 0,
      "inputs": [{"source_ix": 0}],
      "output": "output/$1.o"
    }
  ]
}

command_line_templates describes the commands that compile, transform, link, etc. files. In that example, we only have 1 command. This command is clang++, a C++ compiler. Its arguments are component of several parts. Each part has literal arguments, and variable arguments. The variables are replaced by the correct file names when it tries to update a given file. In the example, if we have a file src/foo.cpp, then upd will instantiate the following command line:

clang++ -c -o output/foo.o -I /usr/local/include src/foo.cpp

source_patterns describes what are the source files of the project. In this example, only the C++ source files are considered. upd crawls the filesystem to find all the files matching each of the patterns. Each pattern can specify capture groups using parentheses. For example src/(*).cpp captures the basename of the file without extensions. For src/foo.cpp, it would be foo. These capture groups can be refered to later in rules.

rules describes the relationship from source files to generated files, and between different generated files. It is possible to build a chain of generated files. Each rule need to specify the following:

• command_line_ix: this is the index in base zero of one of the command line templates specified earlier.
• inputs: this is the indices of the source patterns or the rules that give us this rule input files. In this example, a single source pattern is used as input. We could use a rule by using the field rule_ix instead of source_ix.
• output: this is a substitution pattern that indicates the name of the output file for this or these inputs. In that example, it is output/$1.o. $1 refers to the first capture group of the input pattern. It is replaced by the corresponding captured string. For example, if the source pattern is src/(*).cpp and we found a file src/foo.cpp, the captured group is foo. So, the resulting output file name will be output/foo.o.

Contribute

To get started on developing upd:

# install dependencies (alternative: npm install)
# and bootstrap a compiled version of `upd`
yarn

# setup update manifest
./configure.js

# compile `upd` using itself!
dist/upd update dist/upd

# run unit+e2e tests
yarn test