chartletv0.23.0

Screen reader test tasks

The task matrix for testing chartlet output with VoiceOver and NVDA, with the expected result of each task and a place to record what was found.

This is the test plan behind the screen reader rows of the support matrix. No task has been run yet: every result below is “Not run”. A result goes into this page only with the versions it was observed with, and the support matrix changes only after that. An axe pass does not replace these tasks: it checks the markup, not what a person hears.

Setup to record

Fill this in before the first task, once per session.

Item Value
Date, tester
Operating system and version
Browser and version
Screen reader and version
chartlet version (chartlet --version)
Page: files of examples/ opened from a local copy, or the showcase

Pairs to cover: VoiceOver with Safari on macOS, NVDA with Firefox or Chromium on Windows. JAWS stays a named follow-up.

Tasks

Open the .html file of each example (the HTML figure), not only the .svg. A task passes when the expected result holds without sighted help; write down what was announced word for word where it differs.

# Task Example Expected result VoiceOver / Safari NVDA / Firefox
1 Find the chart and read it budget-vs-actual The chart is one image with the title as its name and the generated description (categories, series, highest and lowest value, missing values). The title is announced once. Not run Not run
2 Read the values budget-vs-actual The data table is reachable, announced as a table with its caption, and has a row for every value, including values whose label was left out. Not run Not run
3 Open and close the table disclosure budget-vs-actual The disclosure announces its state (collapsed, expanded) and works with the keyboard alone. Not run Not run
4 Filter series budget-vs-actual The filter is a checkbox per series: reachable by keyboard, announcing name and checked state; hiding a series removes it from the chart and from what is read, showing it again restores both. Not run Not run
5 Dense time series sensor-readings The description gives the number of points, the time span, the series, the highest and lowest value and the missing values; the table is complete and navigable by row and column. The zoom controls (radio buttons) announce their name and state. Not run Not run
6 Flow editorial-dense-flow The description names every step in reading order and says where each leads; the table has a row per step. The focus disclosure (radio buttons) announces its choices. Not run Not run
7 Sequence async-export Participants and the ten messages are named in order of time; the table has a row per message. Not run Not run
8 Architecture and focus booking-containers The description names every component; the focus control is a group of radio buttons that announces its choices, and choosing one keeps the reading order sensible. Not run Not run
9 Dark theme and forced colors budget-vs-actual with "theme": "dark" Nothing is announced differently; in forced colors every mark and label stays visible (note what is lost, forced colors are not yet supported). Not run Not run
10 Zoom to 200 % and 400 % budget-vs-actual No loss of content or function; the table stays reachable. Not run Not run
11 Phone width and keyboard only budget-by-department (mobile variant) Only one of the two variants is read; the hidden one is not announced. All controls are reachable without a pointer. Not run Not run
12 Reduced motion any example with a control No motion that the page does not need when the system asks for reduced motion. Not run Not run

Embedding contexts

Use budget-vs-actual in each context and record what the reader hears and loses.

Context Expected What the host must provide Result
Inline SVG in a page One image with name and description A caption or text next to it if the page wants the numbers Not run
HTML figure (default) Name, description, caption and the data table Nothing further Not run
<img src="chart.svg"> Title and description survive as the image’s text; the figure wrapper, the table and the controls do not apply alt text, because many screen readers ignore the SVG’s own title for an image element Not run
Markdown / README embedding Not yet evaluated alt text, and the numbers in the text if they matter Not run

Recording defects

Log a defect as an issue with the task number, the pair, the versions and the exact announcement. Do not widen a claim in the accessibility guide while a task is open.

Edit this page on GitHub · Docs for v0.23.0