Manual control of axis bounds

[iwbnwif]

This issue is a place holder for the time being and I will add more detail later.

The 'nice' interval calculation is okay for some basic charting, but it is also necessary to have fine grained control over axis bounds and intervals.

Basically, I think the following is needed:

1. Fully automatic range and interval calculation
2. User specified range (max and min values)
3. Automatic calculation of a 'nice' interval given a fixed range
4. Calculation of a 'nice' interval with a user specified number of interval points for a fixed range
5. Calculation of a 'nice' interval with a user specified number of interval points for an automatic range
6. Manual interval setting with an automatic range
7. Manual interval setting with fixed range
8. Setting the number of minor ticks per interval

So it seems the first thing to do is:

1. Calculate the range either automatically or manually
   2a. Either use the user specified manual interval or
   2b. Calculate a nice interval based on a default number of interval points or
   2c. Calculate a nice interval based on a user specified number of interval points
2. Calculate the minor ticks either manual or using the default value.

[psychegr]

Should the default value for ticks create "evenly spaced ticks" regardless of their real value?

[iwbnwif]

Should the default value for ticks create "evenly spaced ticks" regardless of their real value?

The current strategy for number axis is to use 'nice' values as the default intervals, that means one of 1, 2, 5 and 10 or their factors (e.g. 0.1, 0.2, 0.5 and 1.0). The value is also selected to divide the range into approximately 8 intervals.

The range is then adjusted to include the interval below the lowest data point and the interval above the highest.

In my terminology, a 'major tick' is drawn at every interval and these are divided (if needed) by a fixed number of 'minor ticks'. So if the interval represents 0.5 then the minor tick count could be 5, with each minor tick representing 0.1.

I need to override this behaviour for date/time axis (not done yet).

Also, there are a couple of problems:

1. The algorithm can generate different numbers of intervals for different axis. This causes problems on shared axis because it makes the gridlines look weird
2. Small datasets (0 or 1 points) cause real problems!

Finally, I think that it would be useful to be able to match X and Y interval counts or perhaps pixel spacing to facilitate / ensure that the grid 'squares' are actually square. This might be very complicated though!