wtfpython-cn/irrelevant/notebook_generator.py

"""
An inefficient monolithic piece of code that'll generate jupyter notebook
from the projects main README.

PS: If you are a recruiter, please don't judge me by this piece of code. I wrote it
in hurry. I know this is messy and can be simplified, but I don't want to change it
much because it just works. 

Simplifictions and improvements through patches are more than welcome however :)


#TODOs

- CLI arguments for running this thing
- Add it to prepush hook
- Add support for skip comments, to skip examples that are not meant for notebook environment.
- Use templates?
"""

import json
import os
import pprint

fpath = os.path.join(os.path.dirname( __file__ ), '..', 'README.md')
examples = []

# The globals
current_example = 1
sequence_num = 1
current_section_name = ""


STATEMENT_PREFIXES = ["...", ">>> ", "$ "]

HOSTED_NOTEBOOK_INSTRUCTIONS = """

## 托管笔记本指南

这只是通过 jupyter notebook 浏览 wtfpython 的实验性尝试。因为一些示例是只读的：
- 它们要么需要托管运行时不支持的 Python 版本。
- 要么它们无法在此环境中复现。

预期的输出已经存在于代码单元之后的折叠单元中。 Google colab 提供 Python2（2.7）和 Python3（3.6，默认）运行环境。 您可以在这些 Python2 特定示例之间切换。 对于特定于其他次要版本的示例，您可以简单地参考折叠的输出（目前无法在托管笔记本中控制次要版本）。 您可以检查当前的活动版本
```py
>>> import sys
>>> sys.version
# Prints out Python version here.
```

话虽如此，大多数示例都按预期工作。 如果您遇到任何问题，请随时查阅 wtfpython 上的原始内容并在 repo 中创建问题。 祝君顺利！
---
"""


def generate_code_block(statements, output):
    """
    Generates a code block that executes the given statements.

    :param statements: The list of statements to execute.
    :type statements: list(str)
    """
    global sequence_num
    result = {
        "type": "code",
        "sequence_num": sequence_num,
        "statements": statements,
        "output": output
    }
    sequence_num += 1
    return result


def generate_markdown_block(lines):
    """
    Generates a markdown block from a list of lines.
    """
    global sequence_num
    result = {
        "type": "markdown",
        "sequence_num": sequence_num,
        "value": lines
    }
    sequence_num += 1
    return result


def is_interactive_statement(line):
    for prefix in STATEMENT_PREFIXES:
        if line.lstrip().startswith(prefix):
            return True
    return False


def parse_example_parts(lines, title, current_line):
    """
    Parse the given lines and return a dictionary with two keys:
    build_up, which contains all the text before an H4 (explanation) is encountered,
    and
    explanation, which contains all the text after build_up until --- or another H3 is encountered.
    """
    parts = {
        "build_up": [],
        "explanation": []
    }
    content = [title]
    statements_so_far = []
    output_so_far = []
    next_line = current_line
    # store build_up till an H4 (explanation) is encountered
    while not (next_line.startswith("#### ")or next_line.startswith('---')):
        # Watching out for the snippets
        if next_line.startswith("```py"):
            # It's a snippet, whatever found until now is text
            is_interactive = False
            output_encountered = False
            if content:
                parts["build_up"].append(generate_markdown_block(content))
                content = []

            next_line = next(lines)

            while not next_line.startswith("```"):
                if is_interactive_statement(next_line):
                    is_interactive = True
                    if (output_so_far):
                        parts["build_up"].append(generate_code_block(statements_so_far, output_so_far))
                        statements_so_far, output_so_far = [], []
                    statements_so_far.append(next_line)
                else:
                    # can be either output or normal code
                    if is_interactive:
                        output_so_far.append(next_line)
                    elif output_encountered:
                        output_so_far.append(next_line)
                    else:
                        statements_so_far.append(next_line)
                next_line = next(lines)

            # Snippet is over
            parts["build_up"].append(generate_code_block(statements_so_far, output_so_far))
            statements_so_far, output_so_far = [], []
            next_line = next(lines)
        else:
            # It's a text, go on.
            content.append(next_line)
            next_line = next(lines)

    # Explanation encountered, save any content till now (if any)
    if content:
        parts["build_up"].append(generate_markdown_block(content))

    # Reset stuff
    content = []
    statements_so_far, output_so_far = [], []

    # store lines again until --- or another H3 is encountered
    while not (next_line.startswith("---") or
               next_line.startswith("### ")):
        if next_line.lstrip().startswith("```py"):
            # It's a snippet, whatever found until now is text
            is_interactive = False
            if content:
                parts["explanation"].append(generate_markdown_block(content))
                content = []

            next_line = next(lines)

            while not next_line.lstrip().startswith("```"):
                if is_interactive_statement(next_line):
                    is_interactive = True
                    if (output_so_far):
                        parts["explanation"].append(generate_code_block(statements_so_far, output_so_far))
                        statements_so_far, output_so_far = [], []
                    statements_so_far.append(next_line)
                else:
                    # can be either output or normal code
                    if is_interactive:
                        output_so_far.append(next_line)
                    else:
                        statements_so_far.append(next_line)
                next_line = next(lines)

            # Snippet is over
            parts["explanation"].append(generate_code_block(statements_so_far, output_so_far))
            statements_so_far, output_so_far = [], []
            next_line = next(lines)
        else:
            # It's a text, go on.
            content.append(next_line)
            next_line = next(lines)

    # All done
    if content:
        parts["explanation"].append(generate_markdown_block(content))

    return next_line, parts


def remove_from_beginning(tokens, line):
    for token in tokens:
        if line.lstrip().startswith(token):
            line = line.replace(token, "")
    return line


def inspect_and_sanitize_code_lines(lines):
    """
    Remove lines from the beginning of a code block that are not statements.

    :param lines: A list of strings, each representing a line in the code block.
    :returns is_print_present, sanitized_lines: A boolean indicating whether print was present in the original code and a list of strings representing
    sanitized lines.  The latter may be an empty list if all input lines were removed as comments or whitespace (and thus did not contain any statements).
    This function does not remove blank lines at the end of `lines`.
    """
    tokens_to_remove = STATEMENT_PREFIXES
    result = []
    is_print_present = False
    for line in lines:
        line = remove_from_beginning(tokens_to_remove, line)
        if line.startswith("print ") or line.startswith("print("):
            is_print_present = True
        result.append(line)
    return is_print_present, result


def convert_to_cells(cell_contents, read_only):
    """
    Converts a list of dictionaries containing markdown and code cells into a Jupyter notebook.

    :param cell_contents: A list of dictionaries, each
    dictionary representing either a markdown or code cell. Each dictionary should have the following keys: "type", which is either "markdown" or "code",
    and "value". The value for type = 'markdown' is the content as string, whereas the value for type = 'code' is another dictionary with two keys,
    statements and output. The statements key contains all lines in between ```py\n``` (including) until ```\n```, while output contains all lines after
    ```.output py\n```. 
    :type cell_contents: List[Dict]

        :param read_only (optional): If True then only print outputs are included in converted
    cells. Default False
        :type read_only (optional): bool

        :returns A Jupyter notebook containing all cells from input parameter `cell_contents`.
    Each converted cell has metadata attribute collapsed set to true if it's code-cell otherwise None if it's markdow-cell.
    """
    cells = []
    for stuff in cell_contents:
        if stuff["type"] == "markdown":
            # todo add metadata later
            cells.append(
                {
                    "cell_type": "markdown",
                    "metadata": {},
                    "source": stuff["value"]
                }
            )
        elif stuff["type"] == "code":
            if read_only:
                # Skip read only
                # TODO: Fix
                cells.append(
                {
                    "cell_type": "markdown",
                    "metadata": {},
                    "source": ["```py\n"] + stuff["statements"] + ["```\n"] + ["```py\n"] + stuff['output'] + ["```\n"]
                    }
                )
                continue

            is_print_present, sanitized_code = inspect_and_sanitize_code_lines(stuff["statements"])
            if is_print_present:
                cells.append(
                    {
                        "cell_type": "code",
                        "metadata": {
                            "collapsed": True,

                        },
                        "execution_count": None,
                        "outputs": [{
                            "name": "stdout",
                            "output_type": "stream",
                            "text": stuff["output"]
                        }],
                        "source": sanitized_code
                    }
                )
            else:
                cells.append(
                    {
                        "cell_type": "code",
                        "execution_count": None,
                        "metadata": {
                            "collapsed": True
                        },
                        "outputs": [{
                            "data": {
                                "text/plain": stuff["output"]
                            },
                            "output_type": "execute_result",
                            "metadata": {},
                            "execution_count": None
                        }],
                        "source": sanitized_code
                    }
                )

    return cells


def convert_to_notebook(pre_examples_content, parsed_json, post_examples_content):
    """
    Convert a JSON file containing the examples to a Jupyter Notebook.
    """
    result = {
        "cells": [],
        "metadata": {},
        "nbformat": 4,
        "nbformat_minor": 2
    }

    notebook_path = "wtf.ipynb"

    result["cells"] += convert_to_cells([generate_markdown_block(pre_examples_content)], False)

    for example in parsed_json:
        parts = example["parts"]
        build_up = parts.get("build_up")
        explanation = parts.get("explanation")
        read_only = example.get("read_only")

        if build_up:
            result["cells"] += convert_to_cells(build_up, read_only)

        if explanation:
            result["cells"] += convert_to_cells(explanation, read_only)

    result["cells"] += convert_to_cells([generate_markdown_block(post_examples_content)], False)

    #pprint.pprint(result, indent=2)
    with open(notebook_path, "w") as f:
        json.dump(result, f, indent=2)


with open(fpath, 'r+', encoding="utf-8") as f:
    lines = iter(f.readlines())
    line = next(lines)
    result = []
    pre_examples_phase = True
    pre_stuff = []
    post_stuff = []
    try:
        while True:
            if line.startswith("## "):
                pre_examples_phase = False
                # A section is encountered
                current_section_name = line.replace("## ", "").strip()
                section_text = []
                line = next(lines)
                # Until a new section is encountered
                while not (line.startswith("## ") or line.startswith("# ")):
                    # check if it's a H3
                    if line.startswith("### "):
                        # An example is encountered
                        title_line = line
                        line = next(lines)
                        read_only = False
                        while line.strip() == "" or line.startswith('<!--'):
                            #TODO: Capture example ID here using regex.
                            if '<!-- read-only -->' in line:
                                read_only = True
                            line = next(lines)

                        example_details = {
                            "id": current_example,
                            "title": title_line.replace("### ", ""),
                            "section": current_section_name,
                            "read_only": read_only
                        }
                        line, example_details["parts"] = parse_example_parts(lines, title_line, line)
                        result.append(example_details)
                        current_example += 1
                    else:
                        section_text.append(line)
                        line = next(lines)
            else:
                if pre_examples_phase:
                    pre_stuff.append(line)
                else:
                    post_stuff.append(line)
                line = next(lines)

    except StopIteration as e:
        #pprint.pprint(result, indent=2)
        pre_stuff.append(HOSTED_NOTEBOOK_INSTRUCTIONS)
        result.sort(key = lambda x: x["read_only"])
        convert_to_notebook(pre_stuff, result, post_stuff)