Testing Swage — here's where to look

What's worth trying

• Your own DWG from real work -- not a test file, with your own layers, naming and edit history.
• Write to several formats in one job at once -- compare what each kept and what it didn't.
• A file with no coordinate system -- see how we name that and what we offer instead.
• A big file -- thousands or tens of thousands of features.
• An archive: several files, or a drawing with its side files together, one ZIP.
• If you work with roads -- pick the "roads only" profile on upload and check the class in by-class.gpkg against what you can see on the drawing yourself.

What's already known not to work

• No accounts and no payment: usable with no sign-in, free.
• The processing schema can't be edited by mouse -- only chosen from ready-made schemas.
• Your own transformers and plugins can't be plugged in.
• Publishing your own schema for someone else's use isn't possible.
• We don't read DGN version 8 -- no free driver exists for it in any build.
• Some DWG files don't survive the free converter: it hands back a fragment instead of the file. We catch this and call it a refusal, but we cannot read it today.
• We don't unpack 7z and RAR archives.
• MapInfo TAB and Shapefile don't keep elevation: writing to these formats drops the Z value, and this is shown as a number on the result screen.
• KML only holds a coordinate in WGS84 degrees -- a file with no coordinate system will not write to KML.

The limits in effect right now

• A file -- up to 300 MB.
• The result is kept 24 h after it's ready, then deleted along with everything made from it.
• No more than 12 uploads per hour from one address.
• No more than 3 jobs at once from one address.
• The stand's whole queue -- no more than 20 jobs at a time.
• No more than 434110 objects in one file. We refuse right away above that.
• We read DWG up to 345 MB, DXF and tables up to 1035 MB. Above that we refuse before parsing.

What happens to your data

We do not store your data. The file is deleted as soon as it is no longer needed for processing. What we read — our model of the file — is deleted once the processing period ends (24 h). The final file exists only with you: we assemble it from the model at the moment of download, and keep it available for download no longer than 24 h. What stays with us afterwards is anonymized data only — statistics and the outcome of processing, counters, the data's schema — tied neither to the source file nor to you. Neither the source nor anything derived from it is passed to anyone. Among that anonymized data we also keep a salted fingerprint of your dataset's layer names, used only to notice when unrelated files share the same schema.

Where to write

The "Ask us" button -- bottom right on any page.

On your job's own page -- the "Something is wrong with this job — tell us" button: the job number is already filled in, no need to look for it.