runtime error 201 using Fit.ADDXY on over 10K points lazarus / 55-TAChart ---------------------------------------------------------------------- Ruptor · 2013-02-02 22:49 If I plot 15000 points with Line.ADDXY I don't get a problem but Fit.ADDXY gives error 201 somewhere between 10000 and 15000 points on the graph. My thinking is that if Line works then Fit should too but I don't know enough about it. ---------------------------------------------------------------------- Ask · 2013-02-02 23:01 I tried to reproduce as follows: 1) Open "fit" demo 2) Change "N = 100" at line 169 of Main.pas to "N = 20000" 3) Run, try various fit parameters I did not experience any errors. Please post a small application demonstrating your problem. ---------------------------------------------------------------------- Ruptor · 2013-02-02 23:46 Here iis my test project. You will have to change the directory to where you put the the Dump50K data file. Just click the file and run. For some reason it goes wrong after more points on my home computer but less than 15K on my other computer. ---------------------------------------------------------------------- Ask · 2013-02-03 01:22 Funny, the actual limit turned out to be 20000 -- the exact number I checked. Point number 20001 indeed causes range check -- I missed it by just one. This is a bug/limitation of numlib library, which I reported to bugracker. Anyway, committed a workaround in r40136 which increases the limit to about 1000000 -- I hope it will suffice for you. ---------------------------------------------------------------------- Ruptor · 2013-02-03 01:52 Why would the problem occur at a lower value <15000 on my other computer? The only difference between them is the debugger with the lower value happening on the 7.2 debugger and the high value on the 7.4 version. ---------------------------------------------------------------------- Ask · 2013-02-03 02:00 Not sure about that -- you can look at the bug report I created to see exact limit value on FPC source. Anyway, please test -- the problem should be fixed in TAChart. ---------------------------------------------------------------------- Ruptor · 2013-02-05 13:53 Sorry to say I tested it in the released 1.0.6 version of Lazarus that says data sets over 100K have been fixed but I still have the problem with 50K data file. The revision says 40119. The problem reports as exception error 201 at line 445 in ipf.pas. ---------------------------------------------------------------------- Ask · 2013-02-05 15:38 No, this fix was not included in 1.0.6 release. ---------------------------------------------------------------------- Ruptor · 2013-02-05 16:15 The release bug fixes states "r40043    TAChart   Fix drawing of line series with number of points above 1000000" Not much point in documentation that is wrong is there? You said it is 100K in r40136 but how do I get that version? Please be patient because I am new to Lazarus and don't know how everythimg works. Thanks. ---------------------------------------------------------------------- Ask · 2013-02-05 21:45 QuotePlease be patientOk. Note the number at the start of the line. It is a revision number, which uniquely identifies change in the Lazarus repository. Also note that 40043 and 40136 are different numbers. If you are interested, r40043 fixes the problem from this topic You did not say which page you were looking at for the change log, I suppose it is the standard one. On that page, you can see sections titled "Fixes for 1.0.8", "Fixes for 1.0.6", etc. Note that 1.0.8 is a future version which was not released yet. If you search for r40136 on this page (to use your browser's searching function, press Ctrl+F), you will notice that it is listed under "Fixes for 1.0.8" section. Quotehow do I get that version?By either building from source or using snapshots. ---------------------------------------------------------------------- Ruptor · 2013-02-06 00:57 I see a snapshot today that has r40186 so I installed it and 50K points works now. I have to say there have been some staggering improvements because my machine only takes 2 seconds to plot the fitted line on 50K points where as before, even on 15K points that did work, the code would slow down and take 30 seconds. Well done and thank you. Lazarus is looking perfect for my task.