I agree that full freedom will not provide some helpfully automatic hierarchical reordering (now is done by the algorithm), but many times it is too strict, not doing what I'd want, and ctrl/shift-drag only help limitedly.
Often I'd want to place children closer, even right next or above to a parent (which also points to another popular requested feature, freeing up the parent-child hierarchy itself) or place them in a circle around a parent. And I'd want to branch out vertically or diagonally, instead of only horizontally. Often, I don't know how big the branch will be so I search for a free spot. (There's plenty up and down from the center node...) But once the branches grow they start pushing each other in all kinds of unwanted directions (as long as you don't understand the rules the algorithm applies) making manual graphical correction a growing task in the process.
The point is, that it's ok to have an algorithm help you order according to a global grid or rule set, but in practice it works best for me when applied to *not all nodes or branches*, allowing for a mix of paradigms (horizontal/vertical, relative density, concentric/linear, etc) and some manual control for exceptions. Some other programs offer magnetic grids and templates, but I run into similar problems because they always act globally, not allowing for the typical hybrids I end up to require. Even when I'm doing data visualisation completely manually (eg. in Illustrator) I end up searching for a global form paradigm (which can be translated to an ordering algorithm) with some exceptions to the rule.
Maybe making a grid & algorithm options visible and limitedly accessible for (local & global) change could already help a lot.
Sytse
I agree that full freedom will not provide some helpfully automatic hierarchical reordering (now is done by the algorithm), but many times it is too strict, not doing what I'd want, and ctrl/shift-drag only help limitedly.
Often I'd want to place children closer, even right next or above to a parent (which also points to another popular requested feature, freeing up the parent-child hierarchy itself) or place them in a circle around a parent. And I'd want to branch out vertically or diagonally, instead of only horizontally. Often, I don't know how big the branch will be so I search for a free spot. (There's plenty up and down from the center node...) But once the branches grow they start pushing each other in all kinds of unwanted directions (as long as you don't understand the rules the algorithm applies) making manual graphical correction a growing task in the process.
The point is, that it's ok to have an algorithm help you order according to a global grid or rule set, but in practice it works best for me when applied to *not all nodes or branches*, allowing for a mix of paradigms (horizontal/vertical, relative density, concentric/linear, etc) and some manual control for exceptions. Some other programs offer magnetic grids and templates, but I run into similar problems because they always act globally, not allowing for the typical hybrids I end up to require. Even when I'm doing data visualisation completely manually (eg. in Illustrator) I end up searching for a global form paradigm (which can be translated to an ordering algorithm) with some exceptions to the rule.
Maybe making a grid & algorithm options visible and limitedly accessible for (local & global) change could already help a lot.