st_length on curve is on linearized geom

[mjurce]

Executing: select st_length(st_geomFromtext('CIRCULARSTRING(0 1000,707.1067812 707.1067812,1000 0)'));

st_length

---

1570.63862548611

Gives same result as: select st_length(st_curvetoline(st_geomFromtext('CIRCULARSTRING(0 1000,707.1067812 707.1067812,1000 0)')));

st_length

---

1570.63862548611

The function should return the arc length and not the approximated linestring length.

[pramsey]

There's lots of stuff in the arc support that doesn't work directly on arcs. Heck, distance didn't work directly on arcs until 2.1. So upgrades for full arc support are enhancements. I'm a little surprised that length isn't implemented, and will verify if that is the case, since it's so easy to do.

[robe]

should this be 2.0.4 or 2.1.1 instead. If it requires a lot of code for 2.0.4 (which doesn't have your new fangled code), we probably shouldn't mess with it in 2.0.4

[pramsey]

Yep, confirmed, length code just segmentizes, right up to trunk.

[robe]

and so? is this fixable in next 2 weeks or punt to 2.2 (or 2.1.2)

[pramsey]

I had the maths worked out, will probably be in this week.

[pramsey]

Fixed in 2.1 at r12043

Fixed in trunk at r12045

[strk]

I noticed compoundcurve is still using the linearized geom for length, should I reopen this ?

[mjurce]

Yes.

[strk]

This being from 4 years ago I guess it's better to file another ticket. Would you do verify my finding, mjurce, and open the new ticket ? What I noticed was in the C code, didn't try to expose it at the SQL level...

[strk]

I've filed #3774 to tackle compoundcurve