Skip to content

Latest commit

 

History

History
170 lines (151 loc) · 5.65 KB

[offbin] Binarise tool.md

File metadata and controls

170 lines (151 loc) · 5.65 KB
Raw

offbin

Purpose

Single command to run any of the commands that were previously in the directory.

Consider

  • Only supporting Makefile, cargo make, mage, package.json &etc., rather than making offbin a generic self-extracting format

CLI arguments for offbin

  • --version
  • --help
  • -v, specify multiple times for greater verbosity (log levels)
  • --default-pipe, specify default for all: --pipe '*:stdout:file:///dev/stdout' --pipe '*:stderr:file:///dev/stderr', and specialisation: package.json:stdout:'file://$name.stdout'
  • --dry-run: don't execute, just show what will be executed
  • --default-arg, -A: argument. Added to the end of every default/specialised entrypoint. Can be specified multiple times.
  • --no-ignore-vcs: rather than ignoring everything in .gitignore, will compress entire folder into the output binary.
  • --no-defaults: no default entrypoints. Handy for using offbin as a simple self-extracting archive format, and for when --entrypoint are specified explicitly but not exhaustively.
  • --default-entrypoint: explicit command to run, rather than default. Can be specified multiple times.
  • All positional refer to tasks, <binary> task0 task1 will run tasks: task0 and task1 sequentiallyf

CLI arguments for offbinifed binary

  • --version
  • --help
  • -v, specify multiple times for greater verbosity (log levels)
  • --pipe, specify default for all: --pipe '*:stdout:file:///dev/stdout' --pipe '*:stderr:file:///dev/stderr', and specialisation: package.json:stdout:'file://$name.stdout'
  • --dry-run: don't execute, just show what will be executed
  • --no-cleanup: don't remove extracted files after run
  • --arg, -A: argument. Added to the end of every default/specialised entrypoint. Can be specified multiple times.
  • --no-defaults: no default entrypoints. Handy for using offbin as a simple self-extracting archive format, and for when --entrypoint are specified explicitly but not exhaustively.
  • --entrypoint: explicit command to run, rather than default. Can be specified multiple times.
  • All positional refer to tasks, <binary> task0 task1 will run tasks: task0 and task1 sequentially

Idea

Example 0

> dir /b
foo.exe foo.toml can.exe can.toml runner.toml
> type runner.toml
[runner]
foo --makefile foo.toml
can --makefile can.toml
> offbin --compression=upx -o ..\cool.exe
> cd ..\tmp_dir
> ..\cool.exe
# Extracting with UPX
# Running foo with arg "foo.toml"
# Running can with arg "can.toml"
# Removing all extracted files
> ..\cool.exe foo
# Extracting with UPX
# Running foo with arg "foo.toml"
# Removing all extracted files

Example 1

Setup

$ ls
package.json  Cargo.toml  Rakefile  Makefile
index.js      main.rs     cli.rb    main.c
> offbin --compression=lzma -o /tmp/multi

0

$ cd "$(mktemp -d)"
$ /tmp/multi
# Extracting with tar
# Running 'package.json' with default `npm run-script build`
# <output from ^>
# Running 'Cargo.toml' with default `cargo run`
# <output from ^>
# Running 'Rakefile' with default `rake`
# <output from ^>
# Running 'Makefile' with default `make`
# <output from ^>
# Removing all extracted files

1

$ cd "$(mktemp -d)"
$ /tmp/multi package.json
# Extracting with tar
# Running 'package.json' with `npm run-script hello`
# <output from ^>
# Removing all extracted files

2

$ cd "$(mktemp -d)"
$ /tmp/multi --entrypoint 'Cargo.toml:ls'
# Extracting with tar
# Running 'package.json' with default `npm run-script build`
# <output from ^>
# Running 'Cargo.toml' with `ls`
package.json  Cargo.toml  Rakefile  Makefile
index.js      main.rs     cli.rb    main.c
# Running 'Rakefile' with default `rake`
# <output from ^>
# Running 'Makefile' with default `make`
# <output from ^>
# Removing all extracted files

3

$ cd "$(mktemp -d)"
$ /tmp/multi --entrypoint 'Cargo.toml:ls' --no-defaults
# Extracting with tar
# Running 'Cargo.toml' with `ls`
package.json  Cargo.toml  Rakefile  Makefile
index.js      main.rs     cli.rb    main.c
# Removing all extracted files

Implementation detail

Because many different language formats will be supported, create a mapping in a language-independent format, like:

{
  "Makefile*": {
    "command": "make $@",
    "name": {"$ref": "#/README.md/name"}
  },
  "Gruntfile": {
    "command": "grunt $@"
  },
  "makers": {
    "command": "makers $@"
  },
  "package.json": {
    "command": "npm run-script build",
    "name": ".name"
  },
  "Makefile.toml": {
    "command": "cargo make $@",
    "name": {"$ref": "#/Cargo.toml/name"}
  },
  "Cargo.toml": {
    "command": "cargo run",
    "name": ".name"
  },
  "README.md": {
    "command": null,
    "name": "get_name_from_README()"
  }
}

Also environment variables MUST be inherited from base process.

Ideal things to package

Focus on technologies which work on 1 machine, e.g.: a single server or dev machine.

Cross platform version management for interpreters:

Potentially distributed solutions:

  • Postgres, MySQL, &etc.
  • Elastic, RabbitMQ, &etc.
  • nginx, httpd, &etc.

Then, for consensus requirements—i.e.: distributed system—use some combination of mechanisms to deploy, e.g.: offbin for installing base + SSH + key-value sharing so confs can be updated with offbin-produced binary. This will likely end up with 2-3 distinct binaries, ssh, etcd (or similar that we custom create for the sole purpose of bootstrapping confs on different nodes, and orchestrating instantiation), and offbin.