What v1.8.0 adds, annotated

The docs site keeps its gallery and its per-chart demos. The app is a page of its own, with its own header and footer, so the chart has the full width of the screen. The header has a tab for your data and a tab for each data domain, saying how many of that domain’s charts the loaded data supports. Beneath the tabs, on one line, are the open domain’s charts.

The demo study, on the hepatic explorer. Each domain has a hue, carried by a small hex beside its name: blue for labs and vitals, violet for ECG, teal for adverse events. The open chart’s chip is highlighted, and that chip is the view’s name: there is no separate heading taking up room.

With a participant selected, the chart’s own controls, the chart and the participant profile sit side by side. Keeping the app’s own parts to a header and a footer is what leaves the chart room here.

• A filled hex means the chart is ready. A red one means something it needs is missing, and the chip says how many things. A hollow one means no file is loaded for it.
• On the demo study all thirteen charts are ready. The Patient Journey Explorer is a prototype and is not in the app.
• A chart’s status now means one thing. Prototype: not ready for production, on the docs site only. Experimental: still being worked on and fine to use, in the app with a badge. No badge: stable.
• The four standard domains, their columns, and what each chart needs from them are on the new Domains page, generated from one manifest that a test holds to every chart’s own schema.

Drop CSV or JSON files on the Data view. Each file is read in the browser and placed in a domain by its column names, with the number of that domain’s columns it carries. A sidebar keeps the work in sight: three steps, each showing where it stands, and the list of loaded files. A test asserts that no network request leaves the page from the first file onward.

The Data view as it opens. The sidebar’s three steps carry their own buttons, and the list beneath them flags a file only when one of its rows wants a look. The demo study menu offers three studies: the pilot study, a renamed-column study, and a labs-only liver cohort. The pilot study’s description says its labs file carries 110 synthetic liver and kidney participants who are in no other file. Chart views have no sidebar.

A study whose columns carry none of the default names, loaded after Reset. Four files are placed; the fifth belongs to no domain, and the page says so in one sentence and keeps it, with a picker to place it by hand. The sidebar counts what is left to do: 23 rows guessed, and six rows that a chart is waiting on.

One table per file: what the charts need on the left, your column on the right. A row filled because the name matched says “same name”. A row filled from a short list of known SDTM and ADaM alternatives says “guessed”. Nothing fuzzier is attempted. An empty row says how many charts need it. The same table asks what your data calls the measures some charts find by name. The sidebar stays in view as the tables scroll, and choosing a file there jumps to its table.

Before any correction, a chart that cannot draw says what it is missing, by name.

After six rows are set by hand, the same chart draws from the user’s own files. The data is never renamed or rewritten: the mapping only tells each chart which column to read.

The same app as one HTML file, under 1 MB, opened from disk with the network off. It starts empty, offers no demo studies, and uses your system’s fonts, because it loads nothing.

A prototype of a fourteenth chart, and the first that looks at one participant and not across them: exposure, dose changes, adverse events, labs, concomitant medications, medical history and disposition as stacked lanes on one study-day axis. Clicking any mark anchors time on it and lists what was recorded around it.

One participant’s course on one study-day axis. Click any event and the panel beside the chart lists what was recorded around it, each item linked to its source row.

When the RhoInc and SafetyGraphics renderers were retired they left 282 open issues behind. A survey of all twelve trackers found 144 still worth carrying over and picked nine that the shared modules made cheap. All nine are in this release.

The Shift Plot on a log scale, asked for in 2016. Both axes share one domain, so the identity line still means no change.

The QT explorer now prints the confidence intervals it was only drawing, and carries its caution in every view.

• A filter means the same thing in every chart: it can open on a value, drop its All option, or take several values at once, and the dropdown always shows what the chart is filtered by. One test runs the same cases against all twelve charts.
• A measures setting chooses which measures the Measure control offers, and in what order.
• A Reset chart control on nine charts returns filters, measures, visits, scales and selections to how the chart opened.
• The Adverse Event Explorer says what kind of empty an empty table is.
• The Shift Plot learns a log scale.
• The QT explorer prints its confidence intervals as a table, and carries its safety language permanently. Replicate readings are averaged per participant and visit first, so the sample size counts participants.
• The Hepatic Safety Explorer can include or exclude unscheduled visits, and re-derives baseline and peak when it does.
• Fixed: the Adverse Event Explorer drew no counts for a single-arm study with the per-group columns switched off.