This tool can generate male threaded objects which can be appended to a user's model, or a positive object used for cutting female threads into a user's model using difference().  Male parts have optional and configurable bores and spigots.  Female parts have optional and configurable ability to cut straight fluid passages through the object to which they are applied. Sizes from 1/8" NPT to 12" NPT are supported.  I have no idea who would print a 12" NPT part, but good luck with that.

This tool roughly adheres to ANSI/ASME B1.20.1-1983, and should run in 2015.03 or newer.

See the included calibration test file. Those files can be used to determine what printer compensation parameter you need to use.  I included a few prerendered example STLs, but any other sizes can be generated from the corresponding SCAD file.  Just test a couple sizes across the range you're likely to use, and everything else should be in the same ballpark.   Using +0.2 for female parts and -0.2 for male parts seems to be a good place to start.  Due to the radius of curvature, you may find that the smallest female parts need slightly more compensation.

For reference, I've included a spreadsheet containing the profile diagrams and dimensions, as well as some relevant derived parameters.  Anything else can be figured out by reading the code and commentary.

I shouldn't have to tell you that using FDM printed parts for pressurized fluid applications is going to be somewhere between a disappointment and a disaster. You have been warned.

I had been using the Dan Kirshner thread library for a long time, and I know a lot of other people use it too -- especially when it comes to pipe threads.  It's nice in some regards, but I wanted something better for the task.   The DK tool constructs the thread from a series of slightly intersecting polyhedra.  The overlap workaround results in thousands of expensive and problematic near-degenerate triangles (5-30% of all triangles).  In contrast, this tool constructs the thread as a single clean polyhedron, directly from vertex data and explicit triangulation.

This tool does not require the user to look up thread tables and guess about depthing, and it should produce cleaner, less expensive geometry.  For a comparable resolution, the DK tool will produce files which are ~4-6x as large, and will take ~10-20x as long to render.  Because models created with the DK tool will have a large number of low-quality triangles, they will be about 10x more likely to contain incidental defects compared to this tool.  These claims are based on about 80 hours worth of automated test renders up to 4" It's definitely not perfect, but it's better than what I was using.

For comparison, a giant 12" NPT male stub, at default resolution (283 segments) will F6 render in under 5 minutes on 16 year old dumpster hardware.  The exported model is 11 MB as STLA (or 3 MB as STLB).  The same part using the DK tool will take 90m to render, exporting to a 65 MB STLA (or 17 MB as STLB).  The expense of using the DK tool makes it impractical to publish a collection of threaded models using it.

Greatly reduced render time and file size is relevant to anyone who wants to upload collections of prerendered STLs.  If I could make OpenSCAD export to STLB, I'd do that too, but it's up to you to fix that yourself.