addfs

type PerNodeFunc ¶

type PerNodeFunc interface {
	Apply(context.Context, fsnode.T) (adds []fsnode.T, _ error)
}

PerNodeFunc computes nodes to add to a directory tree, for example to present alternate views of raw data, expand archive files, etc. It operates on a single node at a time. If it returns any "addition" nodes, ApplyPerNodeFuncs will place them under a sibling directory called "...". For example, suppose we have an input directory:

parent/
└─dir1/
  ├─fileA
  ├─fileB
  └─dir2/

and we call ApplyPerNodeFuncs(parent/, ourFns). The resulting directory tree will be

parent/
├─.../
│ └─dir1/
│   └─[ nodes returned by PerNodeFunc.Apply(_, dir1/) for all ourFns ]
└─dir1/
  ├─.../
  │ ├─fileA/
  │ │ └─[ nodes returned by PerNodeFunc.Apply(_, fileA) for all ourFns ]
  │ ├─fileB/
  │ │ └─[ nodes returned by PerNodeFunc.Apply(_, fileB) for all ourFns ]
  │ └─dir2/
  │   └─[ nodes returned by PerNodeFunc.Apply(_, dir2/) for all ourFns ]
  ├─fileA
  ├─fileB
  └─dir2/
    └─.../

Users browsing this resulting tree can work with just the original files and ourFns won't be invoked. However, they can also navigate into any of the .../s if interested and then use the additional views generated by ourFns. If they're interested in our_view for /path/to/a/file, they just need to prepend .../, like /path/to/a/.../file/our_view. (Perhaps it'd be more intuitive to "append", like /path/to/a/file/our_view, but then the file name would conflict with the view-containing directory.)

Funcs that need to list the children of a fsnode.Parent should be careful: they may want to set an upper limit on number of entries to read, and otherwise default to empty, to avoid performance problems (resulting in bad UX) for very large directories.

Funcs that simply look at filenames and declare derived outputs may want to place their children directly under /.../file/ for convenient access. However, Funcs that are expensive, for example reading some file contents, etc., may want to separate themselves under their own subdirectory, like .../file/func_name/. This lets users browsing the tree "opt-in" to seeing the results of the expensive computation by navigating to .../file/func_name/.

If the input tree has any "..." that conflict with the added ones, the added ones override. The originals will simply not be accessible.

func NewPerNodeFunc(fn func(context.Context, fsnode.T) ([]fsnode.T, error)) PerNodeFunc