Hi Damir and Rene, In the past , when trying to display time series, I hit exactly this problem. Time series have as the running variable for instance business day. (2/16/2000 = 1751137). I solved my problem (as suggested here by Damir) by shifting my data in the time-series displaying module such that the origin coincided with the first data point and adding this shift back when plotting the X axis labels (which are date strings). Eddy > Rene Brun wrote: > > - Damir Buskulic has raised the question of precision in TAxis > > (float). > > Should we promote float to double in TAxis ? If yes, TH1 > > constructors > > must be changed accordingly. > > Hi Rene, Vincent and others, > > For that question, I don't know if the this is the best way to go. > Clearly, you shouldn't need more than a float to describe an axis. Going > to double will probably solve the problem but not in an definite way, > though it's probably necessary. The problem is that all the coordinates > (so the axis) are relative to an origin that may be very far from the > viewing point, so one has to compute differences between big numbers, > which lead to the behaviour we've seen : > > -------------- > | | > |pad, axis...| > | | > | | > x---------------------------------------------------------- > origin 6 6 > 0 10 10 +1 > > A solution should ensure that all the calculations made for drawing the > axis, histos, etc... never lead to such differences. It could be to have > a shift or reference point at the beginning of the pad. All calculations > would be made refering to this point and only at the end would one add, > for plotting numbers, the distance to the origin (add in double > precision obviously). This certainly leads to changes in TAxis but also > in TH1.
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:19 MET