Convert HTML to PDF

The PDF is built from the structure of the HTML, not from its appearance: headings, paragraphs, lists, tables, quotes and preformatted blocks are laid out cleanly, the page title (the title tag) becomes the document title, and a table’s header rows, made of header cells (th), repeat at the top of every page. A table too wide for the page is printed smaller and, if it still doesn’t fit, split into successive tables that repeat the first column; a message then says so. A picture more than twice as tall as the page (a long infographic, a phone screenshot) keeps its width and continues over the next pages; a less tall one is made smaller to fit one page. This works well for an email saved as .html, an export of documentation or notes, or a report generated by software.

Because the site’s arrangement is not reproduced, a web page saved from the browser also brings its menus, sidebars and footers along as text. Its blocks are written in the order the page shows them on screen (flex boxes, grids) and set apart as it sets them apart, and a table made with styles stays a table; a block the page places at the top of its box, after the start of its first line (a price at the top right of a product card), is written after that line; another block the page places elsewhere by its position (absolute or fixed position) stays where it is in the file, and a message names it. What the page itself hides, with the hidden attribute, with styles written in the file, in a closed dialog or popover or with a fixed-size box that cuts off what overflows it, stays hidden as in a desktop browser, on screen (rules meant for printing hide and show nothing; a message quotes what the printed page shows differently), and a message quotes the text cut off; a line a box cuts in part is kept whole, and a message quotes what the page hides of it; the content of a collapsed section is kept unfolded, and text the page makes invisible without removing it (transparent, zero-sized, the colour of the background, off the screen, in a shape, under a mask or under a filter) is kept: a message quotes both. The bold, italics or strikethrough those styles give are kept, as are the list signs and numbers they set (A., i., ✔, a number forced on an item), the letter case they impose (upper or lower case), applied in the text’s language, and the line breaks and spaces they keep (the white-space property: a Notion export, code pasted from an editor). Small capitals and mirrored, upside-down, vertical or turned text are written as ordinary letters on horizontal lines, and text shown in a forced order (the bdo element) is kept with Unicode control characters that keep that order; a message points out these texts and quotes examples of them. A picture the page turns or flips, itself or through the box it sits in, keeps its angle, and a picture its box crops or surrounds with space (object-fit) is drawn as the page shows it. What a separate stylesheet was hiding, such as a cookie banner, may show up again: a message names those stylesheets, which the converter cannot read; and if the page’s layout cannot be worked out, a message warns that text the page cuts off or hides may be in the file. Lists in a page exported from Google Docs or Word keep their levels and numbering, and symbols in the Symbol and Wingdings fonts of a page saved by Word become the characters they display (α, Ω, ✓). Content added by JavaScript, common on modern sites, is missing from the file. A page made only of frames (such as the index of old Javadoc documentation) is refused, with the names of the files it shows: convert those instead. Images saved in a separate folder or on the web are not included: their description, when they have one, is written in their place, as in a browser that cannot display them, and a message points them out. Page breaks asked for printing (page-break styles, sections of a Word web page) are kept, as are the page size and orientation (@page): if you force A4 or Letter in the settings, only the orientation is kept, and without an @page rule, the pages are A4. An .xhtml file is read the way the browser opens it.