> Alex T. wrote: > For the area calculation, it is actually marginally faster without the > "delete line 1 of p" - i.e. as Geoff Canyon suggested. > ... > So omitting the "delete line 1 of p" is more efficient - as well as > being one line fewer of code.
You are right, of course. I read the thread too fast and overlooked the advice of G.C. It was a silly thing to delete that first line (I'm still a greenhorn). But you are wrong with your last statement, it's only half a line fewer ;-) I tested again also for other improvements. [Secondary result is: LC 6 has at least double speed compared to LC 7-8]. Essential effects had the reduction of the "item-getting" in the "old" function (cited below). It's surprising for the area-only computation. (I also removed the rounding. It's better do have the method of rounding still available.) (A param "@p" gave here no improvement, tested for large p.) -- Centroid and area of a non-intersecting, oriented and closed polygon. -- See https://en.wikipedia.org/wiki/Polygon#Area_and_centroid -- The following function claims to be the fastest for computing centroid (and area): function polyCentroidandArea2 p local A=0, x=0, y=0 put line 1 of p into L1 put item 1 of L1 into x0; put item 2 of L1 into y0 repeat for each line L in p put item 1 of L into x1; put item 2 of L into y1 put x0 * y1 - x1 * y0 into t add (x0 + x1) * t to x; add (y0 + y1) * t to y add t to A; put x1 into x0; put y1 into y0 end repeat if A is 0 then return false else return ( true, x/(3*A), y/(3*A) , abs(A/2) ) end polyCentroidandArea2 -- usage: -- put polyCentroidandArea2(p) into c0 -- if (item 1 of c0) then -- put item 2 to 3 of c0 into myCentroid -- put item 4 of c0 into myArea -- else put "<yourFourLetterWord>" into fld "INFO" -- Based on a technique by Alex T., Geoff C. and Roger G. -- Area of a non-intersecting, oriented and closed polygon: -- The following function claims to be the fastest for computing the area: function polyArea2 p local A=0 put line 1 of p into L1 put item 1 of L1 into x0; put item 2 of L1 into y0 repeat for each line L in p put item 1 of L into x1; put item 2 of L into y1 add x0 * y1 - x1 * y0 to A put x1 into x0; put y1 into y0 end repeat return abs(A/2) end polyArea2 SPEEDTEST with a large non-selfintersecting closed polygon [The testpoints/testgrahic can be generated by command "testPoints"] 32001 points/lines, 10 repeats. The values "x/y" denote x=millisecs per computation with the new method above y=millisecs per computation with the old method below LC 6.7.8-rc3 centroidAndArea 41/46 ( areaOnly 22/27 ) LC 7.1.1-rc3 centroidAndArea 92/155 ( areaOnly 70/110 ) LC 8.0.0-dp8 centroidAndArea 92/155 ( areaOnly 70/110 ) These were the "old" functions. > function polyCentroidandArea p > put line 1 of p into K > put 0 into A; put 0 into x; put 0 into y > repeat for each line L in p > put (item 1 of K) * (item 2 of L) - (item 1 of L) * (item 2 of K) into t > add ((item 1 of K) + (item 1 of L)) * t to x > add ((item 2 of K) + (item 2 of L)) * t to y > add t to A; put L into K > end repeat > if A is 0 then return "Area is zero" > else return ( round(x/(3*A)) , round(y/(3*A)) ) & cr & abs(A/2) > end polyCentroidandArea > function polyArea p > put line 1 of p into K; delete line 1 of p > repeat for each line L in p > add (item 1 of K) * (item 2 of L) - (item 1 of L) * (item 2 of K) to A > put L into K > end repeat > return abs(A/2) > end polyArea -- Command creates a polygon with testpoints, deviating randomly -- its y-coordonates from an approximate ellipse. -- n0 = num of points, r0 = y-radius, r2 = x-radius, lc = centerpoint on testPoints lc local n0=32000, r0=128, r1=10000 --> adjust these values if lc is empty then put 0 into x; put 0 into y else put item 1 of lc into x; put item 2 of lc into y end if if there is no grc "poly" then create grc "poly" set style of grc "poly" to "polygon"; set lineSize of grc "poly" to 2 set foreColor of grc "poly" to (255,255,0) set editmode of grc "poly" to "polygon" --> great feature of LC set randomseed to 1447250307 --> now we get always the same points generated put 2*pi/n0 into cn; put empty into pts repeat with j= 0 to n0 put cr & ( x + r1*cos(j*cn), \ y + (r0 + random(64) - (j mod 2)*random(128))*sin(j*cn) ) after pts end repeat delete char 1 of pts -- is cr set points of grc "poly" to pts --> LC engine rounds for us end testPoints Note: There arises a lot of identical points or vertical up/down-movements due to rounding when the points of the graphic are set. This has no influence on the results above (a little bit on speed only). ===== Love is just a four-letter-word. _______________________________________________ use-livecode mailing list use-livecode@lists.runrev.com Please visit this url to subscribe, unsubscribe and manage your subscription preferences: http://lists.runrev.com/mailman/listinfo/use-livecode
