Offer rendered height and width on text objects - #425
Conversation
… units rw and rh were computed in raw PDF text-space units rather than viewport pixels, making them ~1.5x too small relative to x/y coordinates. In showText(), canvasWidth, curFontSize, and renderedHeight are all accumulated in PDF text-space (before the viewport transform). x/y come from getCoords_(0,0) which applies the full cumulative canvas matrix (viewport × text matrix), so they ARE in viewport pixels. Fix: multiply the PDF-space components by the viewport scale factor (derived from this.baseTransform) before passing to fillText, so that rw and rh end up in the same coordinate system as x and y. Result for a 20pt×20pt box at (20pt, 20pt): Before: x=1, y=1.199, rw=0.799, rh=0.737 After: x=1, y=1.199, rw=1.198, rh=1.043 (rw≈y, rh≈x as expected) Applied to both the regular-font and Type3-font rendering paths. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
modesty
left a comment
modesty
left a comment
There was a problem hiding this comment.
changes to lib/pdfcanvas.js should comply with W3C/WHATWG Canvas API contract. See inline comments for more details.
The preferred way is to use a private extraction helper or context state instead of extending the public Canvas methods.
| } | ||
|
|
||
| strokeText(text, x, y, maxWidth) { | ||
| strokeText(text, x, y, maxWidth, fontSize, renderedHeight) { |
There was a problem hiding this comment.
fillText and strokeText violate the Canvas API contract.
The W3C/WHATWG Canvas contract defines only text, x, y, optional maxWidth; fontSize and renderedHeight
have no API meaning. The current WHATWG Canvas specification
(https://html.spec.whatwg.org/multipage/canvas.html#dom-context-2d-filltext) and W3C 2D Context reference
(https://www.w3.org/2015/06/2dcontext.html) agree. Extra JavaScript arguments may be passed, but the
standard operation does not consume them; Web IDL
(https://webidl.spec.whatwg.org/#overload-resolution-algorithm) ignores trailing arguments beyond the
declared overload.
2 participants
This pull request introduces improvements to the PDF text rendering pipeline, ensuring more accurate vertical text metrics and enhancing the output data structure. The main changes involve passing and calculating the rendered text height (
renderedHeight) throughout the rendering and data extraction layers, updating method signatures, and reflecting these changes in the output and documentation. Additionally, a new test script is added for verifying hidden field extraction.Rendering and Text Metrics Improvements:
CanvasGraphicsrendering logic inbase/display/canvas.jsto calculate and passrenderedHeightwhen rendering both Type3 and regular fonts, and to compute vertical and horizontal scaling factors for more accurate text box metrics. [1] [2] [3]CanvasRenderingContext2D_class inlib/pdfcanvas.jsto accept and propagate the newrenderedHeightparameter infillTextandstrokeTextmethods. [1] [2]PDFFontclass inlib/pdffont.jsto receiverenderedHeight, calculate horizontal scaling, and output both rendered width (rw) and rendered height (rh) in the text block data. [1] [2]Output Data Structure and Documentation:
readme.mdto describe the newh(rendered text box height) field in theTextsarray, clarifying its meaning and independence from horizontal metrics.Testing:
test/test-hidden-field.cjsto verify extraction and display of text, field, and boxset metrics, including the new rendered width and height fields.