Accuracy

The project

Accuracy

How closely does the browser follow TEX? To find out, a document is typeset by LuaTeX twice from one run: as a pageless PDF – one page as tall as the document, so that nothing is moved to another page – and in the browser by ReflowTEX, with its column exactly as wide as the PDF’s. The two are then compared in two ways, each with a tool to see the results.

Where each glyph is

The first comparison uses no pixels. The position of every glyph and rule in the PDF is read from the PDF, and the position the browser gave it from the page, and the two are matched. This is what the render tests do on every change to ReflowTEX: a short document for each feature (paragraphs, inline mathematics, displays, alignments, lists, footnotes, microtype, rules, TikZ pictures) and all of the AMS sample paper testmath.tex. The short documents are also typeset 100pt narrower and 85pt wider, and compared with the page as the browser breaks it again for those widths. In the short documents no glyph is more than 0.014pt from where TEX put it; in testmath.tex, no more than 0.1pt.

Render tests opens every test: the PDF beside the page, anything that does not match marked on both, and the inspector open on the page. Tools, Render tests says more.

How it looks

The second comparison draws testmath.tex both ways, as a PDF renderer and as the browser draw it, and compares the two pictures pixel by pixel, line by line. As the glyphs are in the same places, what differs is the drawing: the two smooth the edges of glyphs differently, and the PDF renderer puts each glyph on a whole pixel row.

Pixel comparison shows the pictures for the whole paper, with the rows that differ marked.