Internal: document BoxRenderers and the box layout algorithm

This commit is contained in:
Jonathan Daugherty 2015-07-21 19:10:11 -07:00
parent 82a5f72bb2
commit a298955a60

View File

@ -377,6 +377,16 @@ hBox :: [Widget] -> Widget
hBox [] = emptyWidget
hBox pairs = renderBox hBoxRenderer pairs
-- | The process of rendering widgets in a box layout is exactly the
-- same except for the dimension under consideration (width vs. height),
-- in which case all of the same operations that consider one dimension
-- in the layout algorithm need to be switched to consider the other.
-- Because of this we fill a BoxRenderer with all of the functions
-- needed to consider the "primary" dimension (e.g. vertical if the
-- box layout is vertical) as well as the "secondary" dimension (e.g.
-- horizontal if the box layout is vertical). Doing this permits us to
-- have one implementation for box layout and parameterizing on the
-- orientation of all of the operations.
data BoxRenderer =
BoxRenderer { contextPrimary :: Lens' Context Int
, contextSecondary :: Lens' Context Int
@ -422,6 +432,53 @@ hBoxRenderer =
in V.vertCat [img, p]
}
-- | Render a series of widgets in a box layout in the order given.
--
-- The growth policy of a box layout is the most unrestricted of the
-- growth policies of the widgets it contains, so to determine the hSize
-- and vSize of the box we just take the maximum (using the Ord instance
-- for Size) of all of the widgets to be rendered in the box.
--
-- Then the box layout algorithm proceeds as follows. We'll use
-- the vertical case to concretely describe the algorithm, but the
-- horizontal case can be envisioned just by exchanging all
-- "vertical"/"horizontal" and "rows"/"columns", etc., in the
-- description.
--
-- The growth policies of the child widgets determine the order in which
-- they are rendered, i.e., the order in which space in the box is
-- allocated to widgets as the algorithm proceeds. This is because order
-- matters: if we render greedy widgets first, there will be no space
-- left for non-greedy ones.
--
-- So we render all widgets with size 'Fixed' in the vertical dimension
-- first. Each is rendered with as much room as the overall box has, but
-- we assume that they will not be greedy and use it all. If they do,
-- maybe it's because the terminal is small and there just isn't enough
-- room to render everything.
--
-- Then the remaining height is distributed evenly amongst all remaining
-- (greedy) widgets and they are rendered in sub-boxes that are as high
-- as this even slice of rows and as wide as the box is permitted to be.
-- We only do this step at all if rendering the non-greedy widgets left
-- us any space, i.e., if there were any rows left.
--
-- After rendering the non-greedy and then greedy widgets, their images
-- are sorted so that they are stored in the order the original widgets
-- were given. All cursor locations and visibility requests in each
-- sub-widget are translated according to the position of the sub-widget
-- in the box.
--
-- All images are padded to be as wide as the widest sub-widget to
-- prevent attribute over-runs. Without this step the attribute used by
-- a sub-widget may continue on in an undesirable fashion until it hits
-- something with a different attribute. To prevent this and to behave
-- in the least surprising way, we pad the image on the right with
-- whitespace using the context's current attribute.
--
-- Finally, the padded images are concatenated together vertically and
-- returned along with the translated cursor positions and visibility
-- requests.
renderBox :: BoxRenderer -> [Widget] -> Widget
renderBox br ws = do
Widget (maximum $ hSize <$> ws) (maximum $ vSize <$> ws) $ do