ADFA-5489: Add a UID-level network traffic page to the metrics carousel

[davidschachterADFA]

Adds a UID-level network traffic page to the editor's metrics carousel, replacing the brand-mark placeholder that ADFA-5487 shipped as page 2.

Stacked on #1784 (ADFA-5487), which builds the carousel this page slots into. Review or merge that first; this retargets to stage automatically once it lands.

#	Commit
1	feat: add a UID-level network traffic page to the metrics carousel
2	fix: label the network axis in whole units and rest zero on the baseline

Accounting

UID-level, as decided on the ticket: TrafficStats.getUidRxBytes / getUidTxBytes cover every process sharing the app's UID, so Gradle's downloads are included without any socket tagging — the Gradle Tooling and daemon processes share it. There is deliberately no per-feature breakdown; the only two tagged sockets in the tree are the local documentation web server and the JDWP listener, neither of which is interesting here.

The platform counters are cumulative since boot, so what is plotted is the delta between samples. Three cases the raw counters get wrong, each guarded:

• The first sample only establishes a baseline, or the chart opens with a spike equal to everything transferred since boot.
• A counter that goes backwards (reboot, re-based accounting) records 0 rather than negative traffic.
• TrafficStats.UNSUPPORTED (-1), which some devices return, is detected once and latched, so -1 is never plotted as a byte count.

getUsage() hands out copies under a lock — the renderer reads every entry while the sampler appends. MemoryUsageWatcher's equivalent has that race today and still hands out its live buffers.

The logarithmic axis

Traffic spans orders of magnitude — a few hundred bytes of chatter beside a multi-megabyte download — and a linear axis flattens all but the largest burst onto the baseline. MPAndroidChart has no logarithmic axis, so the plotted value is log10(bytes + 1) and the formatter turns labels back into byte units.

The + 1 floors zero, which is the common sample rather than an edge case: an idle IDE transfers nothing and log10(0) is negative infinity. A zero sample plots at exactly 0.0 and the line stays continuous.

Two label decisions that fell out of looking at it on a device:

• Units are decimal (1 kB = 1000 B). The first cut labelled gridlines 9B / 99B / 999B / 9.8KB, because powers of ten divided by 1024 stop looking like decades. Decimal gives 0 B / 10 B / 100 B / 1 kB, and is the convention for throughput anyway.
• Labels show 10^value, not the exact inverse 10^value - 1, which would read 9 B / 99 B / 999 B. One byte is not worth the confusion; the legend carries the exact current figure. Zero is labelled exactly, since log10(0 + 1) really is 0.

Worth a reviewer's attention

Commit 2 fixes a rendering bug that unit tests asserting axisMinimum/axisMaximum passed straight through: LineDataSet defaults to axisDependency = LEFT while the labelled axis here is the right one, so the series were positioned by the disabled, auto-ranged left axis. The axis was configured correctly; the data simply was not drawn against it. An idle chart drew its zero line halfway up a plot whose baseline read 0 B. Only the device showed it. There is now a test on the axis dependency of both series.

Verification

Pixel 6 Pro (arm64), v8 debug, over wifi with a real Gradle sync:

• Both series track real traffic — ~10 kB/s peaks and byte-level chatter legible on one scale — with idle periods resting on the 0 B baseline and the axis reading 0 B / 10 B / 100 B / 1 kB / 10 kB.
• Swiping to the memory page and back returns the full history, so the page is recycling-safe like the memory one.
• Font scale 1.0 and 2.0 on a cold start (EditorActivityKt declares fontScale in configChanges, so a warm relaunch reports stale geometry), landscape, nothing clipped.
• 16 new tests (7 watcher, 13 renderer); full app suite green.

Follow-up: ADFA-5486 (#1785) builds on this.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QeW3M24yD6HNhEnbuNW7Hz

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review for a one-time review, or @claude review always to subscribe this PR to a review on every future push.

_{Tip: disable this comment in your organization's Code Review settings.}

[Daniel-ADFA]

LGTM but I think we need an indicator ( maybe dot indicator?) that it's a carousel

[davidschachterADFA]

@Daniel-ADFA thanks for the review. The affordance you're asking for exists, one PR up the stack — you couldn't see it from here.

#1787 is based on #1784, so what you reviewed has the chart title and nothing else. #1785 (ADFA-5486) adds a left and a right arrow flanking the title, and they are now the only way to move between pages: swipe paging was removed because it competed with panning a zoomed chart and with the editor's drawer gesture, and lost often enough to feel broken. The arrow at each end dims when there is nowhere further to go, so the ends of the carousel are visible too.

Arrows rather than dots was deliberate. Dots report position; arrows report position and are tappable. The panel is a fixed 248dp, so the vertical space dots would occupy is scarce — that is also why the dots the carousel originally shipped with were dropped in #1784 in favour of the title.

If you would still like dots in addition to the arrows once you have seen #1785, say so and I will add them.