What the agent does
The agent inventories the calls before changing the dependency. In a chart component like tj/d3-heatmap, package.json declares "d3": "^3.5.16" and index.js imports it once and uses it in eight ways: d3.scale.linear four times (the x, y, opacity and colour scales at lines 129 to 139), d3.svg.axis twice (lines 145 and 152, each with .orient('bottom') or .orient('left') plus ticks, tickPadding, tickSize and tickFormat), d3.time.scale once (line 126, when axisType is time), and d3.select, d3.interpolateHcl, d3.min, d3.max and d3.extent. The build is make build, which runs Babel over index.js into build/index.js through the Makefile pulled in from react-fatigue-dev; there is no test suite, and example/index.js is a small React page that renders two charts.
Then it reproduces on the new major and names the message precisely. With D3 7 installed, d3.scale and d3.time are not defined, so the first scale line throws TypeError: Cannot read properties of undefined (reading 'linear') under V8, the wording the search box usually gets paraphrased into. D3's own 4.0 changelog explains it: "every symbol in D3 4.0 now shares a flat namespace rather than the nested one of D3 3.x. For example, d3.scale.linear is now d3.scaleLinear", and "In place of d3.svg.axis and axis.orient, D3 4.0 now provides four constructors for each orientation: d3.axisTop, d3.axisRight, d3.axisBottom, d3.axisLeft. These constructors accept a scale". The agent checks each replacement exists on the installed build: scaleLinear, scaleTime, axisBottom, axisLeft, interpolateHcl, select, min, max and extent all do.
The port is mechanical for the scales and axes and not for the join. d3.scale.linear() becomes d3.scaleLinear(), d3.time.scale() becomes d3.scaleTime(), and d3.svg.axis().orient('bottom').scale(this.x) becomes d3.axisBottom(this.x) with the same ticks, tickPadding, tickSize and tickFormat chained on. The trap is in renderBuckets: the code does col.enter().append('g') and then sets transform on col, the update selection, which in 3.x included the entered nodes. The same changelog says "selection.append no longer merges entering nodes into the update selection; use selection.merge to combine enter and update after a data join", so without a .merge() every new column would be appended with no position. The agent rewrites that join with merge and keeps the exit branch.
Proof without a test suite is a rendered chart. The agent runs make build on the new major, then loads the example page's two charts in the sandbox's browser with the same generated data as before and after, and compares the counts of .column groups and bins and the axis tick labels, attaching both screenshots. It leaves the interpolate: 'basis' default and the React example alone, and lists react-fatigue-dev, pinned to a commit on GitHub, as the build dependency the maintainer may also want to look at.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "Our chart throws \"d3.scale.linear is not a function\" (or `Cannot read properties of undefined (reading 'linear')`) after upgrading D3 past version 3. List every d3.* call in the chart code with its line, map each nested D3 3 name to its flat D3 4+ name, and port them, including axes built with d3.svg.axis and orient. Check every data join for enter/update code that relied on D3 3 merging entered nodes, and add merge where needed. Build the component, render the example before and after with the same data, compare the rendered elements, and open a pull request with the screenshots and the mapping table.",
"resources": [
{
"url": "https://github.com/your-org/your-app"
}
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
A pull request that changes index.js and package.json only: the four scale calls, the time scale, the two axes, and the column join with merge, plus "d3" moved to the current major. The description carries a table of each old name, its line and its new name, quotes the two changelog sentences that justify the rename and the join change, and attaches before and after screenshots of the example page with the element counts the agent compared.
A reviewer checks the join rewrite most closely, since it is the one change that would fail silently rather than throw, confirms the tick formatting options survived on the new axis constructors, and decides whether to keep the Babel build as it is.
Run it on every Dependabot pull request that bumps d3
The break lands with the major bump, so the trigger is the pull request that moves D3 past 3. An automation with a github trigger on pull_request.opened and pull_request.synchronize runs this prompt on each one, with the event appended after the prompt, and pushes the port onto the bump's branch when the build or render fails on d3.scale.
For many internal charts still on D3 3, your server sends this request once per repository, each with its own UUID as request_id so a retry returns the run already started, and reads each run.completed webhook; a key starts at most max_concurrent runs at once and queues the rest.
The use case this job follows, with its automation and trigger: Fix a bug from an alert.
Why run it on Rush?
The rename is the easy half of a D3 3 to 4 port; the data-join change is the half that renders wrong without an error. In a Rush sandbox the agent builds the component, draws the example in a real browser before and after, and keeps the screenshots and element counts, so a reviewer sees the chart rather than trusting a search-and-replace.
The changelog that explains each change is long and the agent quotes the exact sentences for the calls this chart uses, nothing more. The pull request is a mapping table with evidence, which is what a port of a chart library needs to be reviewable.
Frequently asked questions
Why does my browser say `Cannot read properties of undefined` instead of `is not a function`?
Because d3.scale itself no longer exists on D3 4 and later, so reading .linear on it fails before any call. Different engines word that differently; the cause and the fix are the same, d3.scaleLinear.
What replaces d3.svg.axis().orient('bottom')?
d3.axisBottom(scale), with d3.axisLeft, d3.axisTop and d3.axisRight for the other orientations. The constructors take the scale directly, and ticks, tickSize, tickPadding and tickFormat chain on as before.
Why does the chart render empty columns after the rename?
Because D3 4 stopped merging entered nodes into the update selection. Code that appends in enter() and then sets attributes on the update selection positions only the nodes that already existed. The agent adds .merge() after the enter append so the attribute calls reach both.
There are no tests. How is the port checked?
By rendering. The agent builds the component, loads the example page with the same generated data on the old and new D3, counts the column groups, bins and axis ticks, and attaches both screenshots to the pull request.
More D3 jobs
More jobs like this: Fix a bug from an alert
- Fix "database is locked" Errors in a GORM and SQLite App (GORM)
- Fix an ASP.NET Core App Bug From a Datadog Monitor Alert (ASP.NET Core)
- Fix a Django App Bug From a Datadog Monitor Alert (Django)
- Fix an Express API Bug From a Datadog Monitor Alert (Express)
- Fix a FastAPI App Bug From a Datadog Monitor Alert (FastAPI)