Why aren't the locators independently enabled/disabled?

Question

I am wondering why the Enabled setting for the first locator (point1) overrides the Enabled setting for the second locator (point2). Any ideas?

Manipulate[
 Graphics[{ {{Arrow[{{0, 0}, point1}]}}, {{Arrow[{{0, 0}, 
   point2}]}}},
  Axes -> True, PlotRange -> {{-4, 8}, {-4, 8}}],
{{point1, {-1, 3}}, {-4, -4}, {8, 8}, {.1, .1}, Locator, Enabled -> False},
{{point2, {5, -2}}, {-3, -3}, {6, 6}, {.1, .1}, Locator, Enabled -> True}]

One of the locators should be enabled; the other should be disabled. However, as the picture shows, they are both disabled.

I realize I could work around this by using LocatorPane, but I'm curious why this doesn't work for individual locators. Am I overlooking something or is this a bug?

Answer 1

As surmised by @Sjoerd in the question's comments, Manipulate combines all locators into a single LocatorPaneBox. We can see this if we inspect the boxes generated by the Manipulate expression:

PaneBox[PanelBox[DynamicWrapperBox[PaneSelectorBox[{True->GridBox[{{
...
LocatorPaneBox[
  Dynamic[{$CellContext`point1$$,$CellContext`point2$$}],
  ...
  Enabled->False,Enabled->True]
...
FrameMargins->{{5,5},{5,5}}],ImageMargins->0,BaselinePosition->Automatic]

This excerpted snippet shows us two things. First, we can see the single LocatorPaneBox that references both points. Second, we can see that the Enabled option is specified twice. As usual, the first instance takes precedence over the second with the result that both locators are disabled. This second point feels like a bug.

To work around this problem is tricky. We could try to use an explicit LocatorPane, but that does not allow us to specify independent regions or enablement states for each locator. If we try to use two LocatorPane expressions, we run into problems when the outer pane is disabled -- events do not get sent down to the inner pane. It is possible to work around this using elaborate EventHandler interactions, but it is not pretty. I suspect that these problems are what led to the present restrictions in the Manipulate implementation.

In this particular case, a simpler solution is to abandon LocatorPane altogether and to include explicit Locator directives instead:

clip[pt:{x_, y_}, {xMin_, yMin_}, {xMax_, yMax_}, {xStep_, yStep_}] :=
  {Round[Clip[x, {xMin, xMax}], xStep], Round[Clip[y, {yMin, yMax}], yStep]}

Manipulate[
  Graphics[
    { Arrow[{{0, 0}, point1}]
    , Arrow[{{0, 0}, point2}]
    , Locator[Dynamic[point1,(point1=clip[#, {-4, -4}, {8, 8}, {.1, .1}])&], Enabled->False]
    , Locator[Dynamic[point2,(point2=clip[#, {-3, -3}, {6, 6}, {.1, .1}])&], Enabled->True]
    }
  , Axes -> True, PlotRange -> {{-4, 8}, {-4, 8}}
  ]
, {{point1, {-1, 3}}, None}
, {{point2, {5, -2}}, None}
]

Unfortunately, this means that we must implement the locator region constraints ourselves since Locator does not offer that service (LocatorRegion only provides crude clipping). However, this is easier than the event-handling gymnastics that would otherwise be required for LocatorPane.

Answer 2

This is just a possible workaround to overcome the issue (WReach has nicely debunked it) by simulating a Manipulate interface. As always, start by writing a DynamicModule object that is then wrapped in a dummy Manipulate wrapper:

Manipulate@DynamicModule[{point1 = {-1, 3}, point2 = {5, -2}},
  Graphics[{
    Dynamic@Arrow[{{0, 0}, point1}],
    Dynamic@Arrow[{{0, 0}, point2}],
    Locator[Dynamic[point1, (point1 = Clip[#, {-4, 8}]) &], Enabled -> False],
    Locator[Dynamic[point2, (point2 = Clip[#, {-3, 6}]) &], Enabled -> True]
    }, Axes -> True, PlotRange -> {{-4, 8}, {-4, 8}}]
  ]