Skip to content

Latest commit

 

History

History
144 lines (133 loc) · 7.64 KB

TODO.md

File metadata and controls

144 lines (133 loc) · 7.64 KB
Raw

High Priority

  • Use more advanced AST library, not the naive builtin ast module. More high level functionalities to provide.
    • astor
    • astunparse
    • libcst
    • astroid
  • Add unit tests
  • Release 1.0.0 version
  • Deploy pre-commit.
    • Generate README TOC
    • Concatenate TODO.md to README.md
    • Reformat
    • Lint
  • Add docstring to module, function, method, etc. Elaborate on what they really do.
  • Release 1.0.0rc1 release candidate version.
  • Fill in unit tests, then deploy Travis-CI and codecov.io
  • 完善 README
  • Instead of LRU, maybe we should use LFU?, or frecency? Investigate different cache replacement policy/strategy.
  • Pre-commit hook to update absort --help message in README
    • 记得脱敏
  • Fill in requirements*.txt files.
  • Remove @profile, @add_profile_decorator_to_class_methods decorators.
  • Reorder functions in main.py by abstraction levels. From higher abstraction level to lower.
  • Fill more metadata in setup.py

Medium Priority

  • Deploy poetry.
  • stackoverflow 关于 Python 测试列表中是否有重复元素的多种 code snippets 写法
  • 应用到 isort 的代码中去，再运行一波，看看有没有问题
  • 尝试使用 PyPy 来运行，看看会不会加速很大幅度
  • 详细看完 Green Tree Sankes 文档，透彻了解
  • visit_Name 不同 context 在 Green Tree Snake 文档中查看
  • Remember timestamp to ignore unmodified files
  • Use multi-thread/multi-process/asyncIO to accelerate when there are large amounts of input files waiting to be processed.
    • Take inspiration from black tool's source code
    • Take inspiration from autopep8 tool's source code
    • File IO is the most expensive and most suitable for multi-thread.
    • CPU-bound components can be accelerated by multi-processing.
    • IO-bound components can be accelerated by multi-threading or async IO.
  • Try to utilize diff-so-fancy when displaying diff.
  • Should we use ThreadPoolExecutor or ProcessPoolExecutor? Test which one yields better performance.
  • Why does ProcessPoolExecutor perform so much worse than ThreadPoolExecutor? Why?
  • 用 line-profiler 测试 absort 项目中多线程 get_dependency_of_decl 是否有助于提升性能，看性能热点占比百分比来定值确定
  • Entries from running rg TODO across the whole repo.
  • Entries from running rg FIXME across the whole repo.
  • 使用 click 第三方插件来实现 mutually-exclusive options.
  • Use click third-party pluggin to realize double confirmation of the -y/--yes option.

Low Priority

  • How to handle name redefinition.
  • Try README-driven development
  • Use semver and conventional commit message guidelines.
  • More detail in symbol table.
    • More detail in symbol table to distinguish different variable with same identifier.
  • When programming, use Tickey to broadcast keyboard sound.
  • Profile to find performance hotspots. Optimize and accelerate the script.
    • Algorithmic optimization
    • Memoization optimization
    • Rewrite-in-Cython optimization
    • Multi-thread/process optimization
    • Rewrite-in-Generator-Style optimization
    • Caculate and Save for next time optimization
  • Provide a programmatical interface.
  • Can we just use tokenize.detect_encoding to replace autopep8.detect_encoding?
  • Use click's mechanism to specify that two options are mutually exclusive.
  • Stress test against CPython site-packages, see if any bugs are spotted.
  • What's the elegant way to pass bunch of cli parameters around functions?
    • Take inspiration from black tool's source code.
    • Take inspiration from autopep8 tool's source code.
    • It might be a good idea and elegant way to use global object to pass command line arguments around.

Uncategorized

Changelog

  • Rename _env_list to _symtab_stack.
  • Rename GetDependencyVisitor to GetUndefinedVariableVisitor.
  • Consider global and nonlocal.
  • Remove unused imports.
  • Copy Graph.py from Source-Concater reposiroty.
  • Reserve comments and blank lines and whitespace layouts. They are important part of the code style.
    • Reserve top level comments
  • Handle comments. We don't want to lose comments after absort processing a file.
  • Add LICENSE
  • Add setup.py
  • 去除 reformat code
  • visit_ClassDef visit keywords attribute
  • topological sort 的时候去除无关的 graph node，这里很容易导致之后维护的时候出错
  • decl_ids 改成 decl_names
  • 添加 cli option --comment-is-attribute-of-following-declaration
  • add confirmation prompt to the dangerous options, eg. --in-place.
  • Add .isort.cfg
  • As a command line tool that deals with files, we should proceed on failure. Not fail fast and early.
  • Use self-customized ast.get_source_segment implementation to accelerate.
    • Algorithmic optimization
    • Memoization optimization
    • Rewrite-in-Cython optimization
    • Multi-thread/process optimization
  • Remove hierarchy level sort related.
  • It's syntactically correct to have two functions depending on each other. Circle in dependency graph should not be error.
  • Fix the bug of starnge result after absorting isort.main module.
  • 只 transform contiguous block of declarations 不要动其他部分的源码. This is also beneficial to reducing diff size, and optimize performance.
  • Fix the bug: Adding heurstic blank lines at the front should be only applicable to declarations.
  • Collect statistics of failure cases. And print summary digest at the end of execution to show user.
  • Extract absort_files
  • Remove more-itertools dependency
  • Some error related information should be printed to stderr, not stdout.
  • Add click parameter help messages.
  • Use intelligent detect file encoding, instead of hardcoding UTF-8. Take inspiration from the autopep8 cli tool.
  • 在 GitHub 设为 public repo
  • Retain as much original layout as possible to reduce diff size.
  • Add .pylintrc
  • Specify release version when installing, in README.
  • Open issue in https://github.com/python/typeshed about NodeTransformer.generic_visit method's return t
  • Check whether ast.iter_child_nodes yield ordered results?
    • ANSWER: judging from source code, it does yield ordered results.
  • 完善 README
    • 添加 GitHub badges
      • use black code style
      • total line count
      • license
    • add examples to showcase
  • Read about Python namespace and scope topics, read them crystal clear, then modify visitors.py.
  • Critical bugs: visitors.py implementation is severely buggy.
  • The whole implementation in visitors.py is a mess and catastrophic! We need REWRITE IT!!!
    • Read Pylint source code to figure out how to properly detect undefined variables.
  • Rewrite, refactor main function. The body code of main function is right now a bunch of mess. Go beyond executor.map/submit. There are more freedom in interface to use. Examine the async/await keyword, Future, Promise, etc.
    • Take inspiration from black tool's source code.
  • GetUndefinedVariableVisitor slots doesn't seem to work?
  • Fix race condition on writing to stdout/stderr in absort_files()
  • Try to use threading.lock or asyncio.lock to solve race condition on writing to stdout/stderr.
  • Refactor absort_files(). Break larget function into multiple small functions.

TIL

  • Path.name is different from Path.str()
    • Path.name is "A string representing the final path component, excluding the drive and root, if any"
    • Path.str is "The string representation of a path is the raw filesystem path itself (in native form, e.g. with backslashes under Windows), which you can pass to any function taking a file path as a string"