Why Sponsor Oils? | source | all versions | oilshell.org

Oils 0.21.0 Quality

This is a supplement to the main release page.

Test Results

Spec Tests

• OSH Survey. Test OSH with existing shells, and compare their behavior.
  • OSH in C++. The progress of the C++ translation.
• YSH. The legacy-free language.
  • YSH in C++. The progress of the C++ translation.
• Stateful Tests. Tests that use pexpect.

Primary Test Suites

• Gold Tests. Compare OSH with bash (using implicit assertions, not golden output.)
• Wild Tests. Parse and translate thousands of shell scripts with OSH.
• Python Unit Tests.
• C++ Test Coverage measured by Clang.
  • Log Files
• Process Table. Are child processes in the right state for job control?

More Tests

• Smoosh test suite (from mgree/smoosh):
  • smoosh | smoosh-hang
• parse-errors. A list of all parse errors.
  • parse-errors-osh-cpp. With the native binary.
• runtime-errors. A list of shell runtime errors.
• ysh-runtime-errors. YSH runtime errors.
• ysh-parse-errors. YSH parse errors.
• ysh-every-string. String literal stress test.
• ysh-large
• lossless. Test an invariant of the parser.
• osh-usage. Misc tests of the osh binary.
• tools-deps. Tests for a subcommand in progress.
• How many processes does Oils start compared to other shells?
  • syscall/by-code | syscall/by-input
• ysh-ify Tests. Test OSH to YSH translation.

Benchmarks

• Parser. How fast does OSH parse compared to other shells?
• Runtime. How fast does OSH run shell scripts?
• Compute. How fast does OSH run small programs without I/O?
• Build. How long does it take for end users to build Oils? How big is the resulting binary?
• Virtual Memory Baseline. How much memory do shells use at startup?
• mycpp. Compares Python and generated C++ on small examples.
• Memory Management Overhead. How much time do we spend managing memory, compared with the shell interpreter?
  • benchmarks/gc. Stats from the OS and our GC runtime.
  • benchmarks/gc-cachegrind. Stable measurements.
• uftrace. Stable measurements for parsing and runtime.

Metrics

• Lines of source, counted in different ways:
  • overview. The whole Oils repo, organized by type of source file.
  • for-translation. An overview of the "compiler engineer" project.
  • cloc-report. Significant lines of code, as measured by the cloc tool.
• Generated C++ code
  • oils-cpp. The C++ code in the oils-for-unix tarball.
  • preprocessed. How much code is passed to the compiler?
    • cxx-dbg, cxx-opt
  • Binary code size reported by Bloaty. How much code is output by the compiler?
    • overview, symbols

Source Code

These files may help you understand how Oils is implemented, i.e. with domain-specific languages and code generation.

• _gen/frontend/id_kind.asdl_c.h. A list of language elements, used in the lexer and in multiple parsers and evaluators.
• The regex-based lexer uses two stages of code generation:
  • frontend/lexer_def.py | _gen/_tmp/match.re2c-input.h | _gen/frontend/match.re2c.h
• frontend/syntax.asdl. The syntax tree for OSH and YSH.
• ysh/grammar.pgen2. The expression grammar for YSH. In contrast, the OSH parsers are hand-written.

Also see the oilshell/oil repository.

Old

These links describe the CPython / "OVM" build, which should become the "experimental" version of Oils.

OPy / OVM Metrics

• Lines of dependencies:
  • pydeps. Oils code plus the Python standard library.
  • nativedeps. Oils code plus a slice of CPython.
• Bytecode Metrics
  • overview - Compare OPy vs. CPython.
  • oil-with-opy - Oils compiled with OPy.
  • oil-with-cpython - Oils compiled with CPython (for comparison).
  • src-bin-ratio-with-opy - How big is the compiled output?
• OVM / CPython
  • overview - An analysis of GCC's compilation of OVM (a subset of CPython). Bloaty provides the underlying data.
  • cpython-defs/overview - We try to ship as little of CPython as possible, and this is what's left.