ST_LineToCurve gives getPoint4d_p offset error

[robe]

I tried to do this:

SELECT ST_LineToCurve(ST_LineMerge(ST_Collect(ST_CurveToLine(ST_GeomFromEWKT('CIRCULARSTRING(0 0, 1 1, 1 0)')),

ST_GeomFromEWKT('LINESTRING(1 0, 0 1)') ))) ;

And was hoping to get this: COMPOUNDCURVE(CIRCULARSTRING(0 0, 1 1, 1 0),(1 0, 0 1))

But instead got error: getPoint4d_p: point offset out of range

[robe]

Slight update. The above example is not a simple line. This seems to behave more or less as expected when I tested with a simple line.

SELECT ST_AsEWKT(ST_LineToCurve(ST_LineMerge(ST_Collect(ST_GeomFromEWKT('LINESTRING(2 2, 2.5 2.5)'), ST_CurveToLine(ST_GeomFromEWKT('CIRCULARSTRING(2.5 2.5,4.5 2.5, 3.5 3.5)')) ))));

Result: COMPOUNDCURVE((2 2,2.5 2.5),CIRCULARSTRING(2.5 2.5,4.20710678118655 1.79289321881345,3.5 3.5))

[mcayland]

Was this fixed with Mark's last patch? If so, can we close?

ATB,

Mark.

[robe]

No so still open. But like I said - this appears to be only an issue with non-simple lines

[mcayland]

Yeah I see. Removing the outer ST_LineToCurve() function from your example above gives the following:

select st_linetocurve(st_geomfromtext('LINESTRING(0 0,-0.0239315652662951 0.0251361090611227,-0.0466009335008786 0.0514162068286816,-0.0679534922100711 0.0787769822452901,-0.0879378012096791 0.107152520806448,-0.106505716548904 0.136474463354359,-0.123612506493335 0.166672170761126,-0.13921695928762 0.197672894104599,-0.153281482438188 0.229401949926901,-0.165772193276862 0.261782900153418,-0.176659000587176 0.29473773623882,-0.185915677096747 0.328187067096473,-0.193519922661073 0.362050310358527,-0.199453417986539 0.396245886505893,-0.203701868763191 0.430691415400453,-0.206255040100989 0.465303914746028,-0.207106781186548 0.499999999999998,-0.206255040100989 0.534696085253969,-0.203701868763192 0.569308584599544,-0.199453417986539 0.603754113494104,-0.193519922661074 0.63794968964147,-0.185915677096747 0.671812932903525,-0.176659000587177 0.705262263761177,-0.165772193276864 0.738217099846577,-0.15328148243819 0.770598050073094,-0.139216959287623 0.802327105895394,-0.123612506493337 0.833327829238871,-0.106505716548901 0.863525536645645,-0.0879378012096717 0.892847479193563,-0.067953492210058 0.921223017754728,-0.0466009335008591 0.948583793171342,-0.0239315652662687 0.974863890938906,3.39173134022985e-14 1.00000000000003,0.0251361090611646 1.02393156526633,0.0514162068287321 1.04660093350092,0.0787769822453494 1.06795349221012,0.107152520806517 1.08793780120972,0.136474463354438 1.10650571654895,0.166672170761214 1.12361250649338,0.197672894104697 1.13921695928767,0.229401949927009 1.15328148243823,0.261782900153537 1.1657721932769,0.294737736238949 1.17665900058722,0.328187067096612 1.18591567709678,0.362050310358676 1.1935199226611,0.396245886506051 1.19945341798656,0.43069141540062 1.20370186876321,0.465303914746205 1.206255040101,0.500000000000184 1.20710678118655,0.534696085254163 1.20625504010098,0.569308584599746 1.20370186876317,0.603754113494313 1.19945341798651,0.637949689641685 1.19351992266103,0.671812932903745 1.18591567709669,0.705262263761404 1.17665900058711,0.73821709984681 1.16577219327678,0.770598050073331 1.15328148243809,0.8023271058956 1.13921695928753,0.83332782923904 1.12361250649325,0.863525536645775 1.10650571654882,0.892847479193655 1.08793780120961,0.921223017754785 1.06795349221002,0.948583793171366 1.04660093350084,0.9748638909389 1.02393156526628,0.999999999999999 1,1.02393156526627 0.974863890938902,1.04660093350084 0.948583793171368,1.06795349221001 0.921223017754787,1.08793780120961 0.892847479193658,1.10650571654882 0.863525536645777,1.12361250649325 0.833327829239042,1.13921695928752 0.802327105895602,1.15328148243809 0.770598050073334,1.16577219327677 0.738217099846851,1.17665900058708 0.705262263761483,1.18591567709666 0.671812932903865,1.193519922661 0.637949689641846,1.19945341798648 0.603754113494515,1.20370186876315 0.569308584599989,1.20625504010097 0.534696085254447,1.20710678118655 0.500000000000509,1.20625504010102 0.465303914746569,1.20370186876325 0.430691415401024,1.19945341798663 0.396245886506491,1.1935199226612 0.362050310359152,1.18591567709691 0.328187067097122,1.17665900058738 0.29473773623949,1.16577219327711 0.261782900154107,1.15328148243848 0.229401949927606,1.13921695928796 0.197672894105317,1.12361250649372 0.166672170761855,1.10650571654934 0.136474463355095,1.08793780121017 0.107152520807188,1.06795349221062 0.0787769822460299,1.04660093350148 0.0514162068294181,1.02393156526696 0.0251361090618519,1 0,0 1)'));

So I'm not sure I agree with your diagnosis on simple lines. This is another crasher we need to fix for 1.4 then :(

[robe]

Its not a crasher, just gives an error so a bit of a slap on the wrist rather than taking the server down. The other one I mentioned is more serious since that really does crash unless you think these two are related.

Mark, you worry too much. We are more liable to break something more important in the process of trying to fix this. Don't worry be happy :).

[mcayland]

It's equivalent to an assertion. At the very least we need to find out what is happening in this code before we make that call.

ATB,

Mark.

[robe]

This does the same for 1.3 so we need to backpatch to 1.3 to close this out. I'm personally of the opinion that if we are going to be fixing bugs that exist in 1.3 as we should do it at the same time as we patch 1.3 and that we should be foxusing on bugs introduced in 1.4 (this is a long standing one). Which is why I really am not in a rush to fix this right now.

SELECT ST_LineToCurve(ST_LineMerge(ST_Collect(ST_CurveToLine(ST_GeomFromEWKT('CIRCULARSTRING(0 0, 1 1, 1 0)')), ST_GeomFromEWKT('LINESTRING(1 0, 0 1)') ))) ;

[mcayland]

Okay. Looks like the curve conversion code needs to peek one point ahead to determine the type of the next segment, and of course if you end your curve at the end of your set of points then this will fail.

Can you try the attached patch which seems to resolve this for me?

ATB,

Mark.

[mcayland]

Applied to 1.4 as r4269 and trunk as r4270. I don't feel the need to backpatch this to 1.3, given that there are several other curve fixes from Mark that won't get backpatched. I think we'd be better off stating that people who want to use curves should go for 1.4 as a minimum.

ATB,

Mark.

[robe]

fair enough

[robe]

Just tested seems okay.