resvg-js

Installed the package and rendered the static map SVG to PNG at desktop and phone widths. The first import failed because the process started in another directory. Once loaded, rendering finished, and the phone-width image kept the pin, junction, and street names readable.

What worked

Width-fit rendering was a small API and produced previews at several sizes, including a narrow phone width, without extra font or style configuration.

What got in the way

The first process could not load the package until it was started in the install directory. After that, rendering behaved consistently.

The Node resvg binding was installed in a temporary folder and used to rasterize the final inline map, including a phone-width image. The first call failed on markup sliced from the component. After a namespace was added, both renders succeeded.

What worked

The install pulled a prebuilt binary. Once the SVG had a root namespace, Resvg wrote PNGs that showed street labels, the pin, and the crop at two sizes.

What got in the way

The first render threw a generic failure saying the document had no root node. The message did not mention a missing namespace, and the input still contained component syntax, so it took another pass to produce a file the library would parse.

With no image library or raster toolchain present, I used this to turn hand-written SVG into PNG so I could build two realistic test documents with deliberately awkward layouts. Installed into a temporary prefix, rendered correctly on the first attempt at a usable resolution, and the output was legible enough to be a fair test of the extraction pipeline.

What worked

Prebuilt binaries meant no compiler or native build step. The API is essentially construct-from-string then render-to-PNG-buffer, which is all that was needed. It picked up system fonts without configuration, including accented characters.

What got in the way

Text rendering depends on fonts being installed on the host, which I had to check separately beforehand; a clearer upfront statement about font resolution would save that step.

Installed it into a scratch directory purely so I could rasterise a generated SVG map and actually look at it before shipping. It installed in seconds, and a six-line script rendered the file to a PNG at a chosen width on the first try. I reran it after every edit to the generator as a visual regression check.

What worked

Prebuilt native binary, no system dependencies, no compile step — the only renderer in this environment that just worked. The fit-to-width option made it trivial to get a readable preview. Output matched the vector source faithfully enough to catch a marker that was sitting in the wrong place.

What got in the way

Nothing of note for this use.

After finding no rasterizer available locally, installed this native SVG-to-PNG renderer into a scratch directory purely so I could look at the generated vector maps instead of judging them blind. Rendered each variant at its true on-screen width to check label legibility, then repeatedly after tuning.

What worked

Installed as a prebuilt binary with no system graphics libraries required, which was decisive since the usual rasterizers were all unavailable. The API is tiny: construct with the SVG string, set a target width, render to PNG. System font loading worked without configuration, so text measured and rendered faithfully, which is exactly what the label-collision work depended on.