The book was cleaned up once before it was reorganized, and this page records how. The rules below were applied page by page: first what was kept, then how each page was shaped, then how links and images were checked, and last how unread notes were filled in. The table at the end counts what happened to every address.
The book pages were rewritten in place. The table of contents was not edited. Sixty-seven unused files were deleted. Nothing in this book linked to them.
The content itself was protected before any wording changed. Sentences were rewritten into clearer English, but facts, quotes, formulas, links, and images were kept.
With the content fixed, each page was given a shape. Each page was written with sections, subsections, and subsubsections where a part sits inside the level above. A heading was added only to name a block already on the page. Existing sections were ordered so the basics come before anything that depends on them. Each section, subsection, and subsubsection opens with one plain-English sentence saying what that part is about.
The blocks inside a page were then written in the form GitBook renders best. A formula was written as GitBook math, and a passage already in quotation marks was written as a quote. A video that is its own block and still plays was embedded, and a link to another page in this book was written as a page link when the file was exact. Tabs, columns, steppers, cards, expandables, hints, task lists, file blocks, and diagrams were used when the notes already had that shape.
Figures got the same care as text. Every figure has a caption, taken from the words next to the image, or from the heading when those words do not name the picture. An existing caption was rewritten into clearer English. An image copied from the web has a credit line. The credit is a source the notes already name, or the address the file was copied from. Google-hosted images that still loaded were saved into the book's image folder, and pages now point at those saved files.
Dead addresses were handled without losing what the author wrote. A Deprecated links section was added only where a link or an image no longer works, and that section keeps the original wording and the dead address. A replacement was used only when it was the same page, paper, file, or image, or the same article matched by title, by sentences, or by both. Towards Data Science links were searched on Medium, and only the same article was used. A replacement article names its author when the article itself gives that name. A wrong-credit note was corrected only when the right credit was found and checked.
Every change was then verified before moving on. Every rewritten link and every local image was checked again. Each page was compared with its original before the next page was started, and anything lost or added outside the plan was corrected. An address that the edit broke was put back.
The last pass filled gaps the author had left open. An unread note was read, and one plain-English paragraph with the technical wording was added. A TODO that pointed at a readable page or file was summarized the same way. A video marked did-not-watch was summarized from its subtitle text, without downloading the video. A section that is the same subject as another page or heading gets one sentence linking there.
The result can be counted. There are 4018 URL occurrences in the rewritten pages. One address used twice counts twice, and one occurrence can sit in more than one row.
| Count | What it means | Number |
|---|---|---|
| Changed | The address in the file is not the address that was there. | 294 |
| Broken | The address is dead and sits in Deprecated links, with no working replacement in the body. | 437 |
| Fixed | The address was dead or pointed at the old Towards Data Science site, and the body now has a working address for the same thing. | 157 |
| Web searched and found | A search found the same article, and that address is in the body. | 157 |
| Web searched and not found | A search was run and no same article was found. | 167 |
| Paywalled | The page stays in the body and is behind a membership or login wall, such as Medium. | 513 |
| Others | The address was already fine and was left as it was, or it does not fit the rows above. | 3175 |