pretty/analyze_structure.lua

local TABLE_TYPE
do
    local thispath, was_loaded =  ... and select('1', ...):match('.+%.') or ''
    was_loaded, TABLE_TYPE         =  pcall(require, thispath..'table_type')
    assert(was_loaded, '[pretty]: Could not load vital library: table_type')
end

local RESERVED_LUA_WORDS = {
    ['and']      = true,
    ['break']    = true,
    ['do']       = true,
    ['else']     = true,
    ['elseif']   = true,
    ['end']      = true,
    ['false']    = true,
    ['for']      = true,
    ['function'] = true,
    ['if']       = true,
    ['in']       = true,
    ['local']    = true,
    ['nil']      = true,
    ['not']      = true,
    ['or']       = true,
    ['repeat']   = true,
    ['return']   = true,
    ['then']     = true,
    ['true']     = true,
    ['until']    = true,
    ['while']    = true,
    ['goto']     = true,
}

local SIMPLE_VALUE_TYPES = {
    ['nil']      = true,
    ['boolean']  = true,
    ['number']   = true,
}

local SHORT_STRING_MAX_LEN = 7
local MINIMUM_NUMBER_OF_SET_ELEMENTS = 2

--------------------------------------------------------------------------------

local function is_identifier(str)
    -- An identier is defined in the lua reference guide

    return str:match('^[_%a][_%w]*$') and not RESERVED_LUA_WORDS[str]
end

local function get_key_types (t)
	local types = { nr_types = -1 }
	for key, _ in pairs(t) do
		types[type(key)] = true
	end
	--
	for type_name, _ in pairs(types) do
        types.nr_types = types.nr_types + 1
	end
	return types
end

local function get_value_types (t)
    local types = { nr_types = -1 }
	for _, value in pairs(t) do
		types[type(value)] = true
	end
	--
	for type_name, _ in pairs(types) do
        types.nr_types = types.nr_types + 1
	end
	return types
end

local function largest_number_index (t)
	-- Returns the largest number index in t.

	local max_index = 0
	for k,v in pairs(t) do
		if type(k) == 'number' then
            max_index = math.max(max_index, k)
        end
	end
	return max_index
end

local function nr_elements_in_table (t)
    local k, count = nil, -1
    repeat
        k, count = next(t, k), count + 1
    until k == nil
    return count
end

local function nr_elements_in_seq (t)
    assert(type(t) == 'table')

    if getmetatable(t) and getmetatable(t).__index then
        return 0, false -- FIXME: Temporary stopgap for when __index throws an
        -- error. I think we need to clone the numbers part of the table, to
        -- fix this.
    end

    local i, last_elem_i, nr_elems, has_holes  =  0, 0, 0, false
    while i <= last_elem_i + 2 do
        i  =  i + 1
        if t[i] ~= nil then
            last_elem_i, nr_elems  =  i, nr_elems + 1
        else
            has_holes = true
        end
    end
    return nr_elems, has_holes
end

--------------------------------------------------------------------------------

local function contains_only_nice_string_keys (t)
	-- Predicate: Does t contain only string keys which could be used as
	-- identifiers, eg.

    for k, _ in pairs(t) do
        if type(k) ~= 'string' or not is_identifier(k) then
            return false
        end
    end
    return true
end

local function is_set (t)
    -- Predicate: Does t contain only boolean values.
    local value_types = get_value_types(t)
    return value_types.boolean and value_types.nr_types == 1
end

local function is_tabular (t)
    -- Predicate: Does t contain sub-tables with identical substructure.

    -- Quick return if table is empty, or not conctaining only values of type table.
    local value_types = get_value_types(t)
    if not value_types.table or value_types.nr_types ~= 1 then
        return false
    end

    -- Determine keys of first child.
    local children_keys = {}
    for key, _ in pairs(t[next(t)]) do
        children_keys[key] = true
    end

    -- Make sure every other child has exact same sub-structure.
    for _, child in pairs(t) do
        for key, _ in pairs(children_keys) do
            if not child[key] then  return false  end
        end
        for key, _ in pairs(child) do
            if not children_keys[key] then  return false  end
        end
    end

    return children_keys
end

local function has_uniform_structure (t)
    -- TODO: This can probably be more relaxed. Maybe combine string, number and boolean?

    assert(type(t) == 'table')

    -- Find the key and value types.
    local first_key = next(t)
    if first_key == nil then  return true  end
    local key_type, value_type =  type(first_key), type(t[first_key])

    -- Ensure every other key value pair is the same.
    for key, value in pairs(t) do
        if type(key) ~= key_type or type(value) ~= value_type then
            return false
        end
    end

    return true
end

--------------------------------------------------------------------------------

local function get_table_info (t)
	local key_types = get_key_types(t)

	local info = {}
          info.address     = tostring(t):sub(8)
          info.nr_elems    = nr_elements_in_table(t)
          info.seq_elems, info.has_holes  =  nr_elements_in_seq(t)
          info.map_elems   = info.nr_elems - info.seq_elems
	      info.has_seq     = info.seq_elems > 0
	      info.has_map     = info.map_elems > 0
          info.is_set      = is_set(t) and info.nr_elems >= MINIMUM_NUMBER_OF_SET_ELEMENTS
          info.is_tabular  = is_tabular(t)
          info.is_uniform  = has_uniform_structure(t)

    -- Determine type of table
	if not info.has_seq and not info.has_map  then  info.type = TABLE_TYPE.EMPTY
	elseif info.has_seq and not info.has_map  then  info.type = TABLE_TYPE.SEQUENCE
    elseif info.is_set                        then  info.type = TABLE_TYPE.SET
    elseif info.has_seq                       then  info.type = TABLE_TYPE.MIXED
  	elseif contains_only_nice_string_keys(t)  then  info.type = TABLE_TYPE.STRING_MAP
    else                                            info.type = TABLE_TYPE.PURE_MAP
  	end

	return info
end

--------------------------------------------------------------------------------

local function analyze_structure (root, max_depth)
    if type(root) ~= 'table' then  return {}  end
    assert(type(root) == 'table')
    local max_depth = max_depth or math.huge
    assert(type(max_depth) == 'number')

    local info, visited, next_mark, depth  =  { root = root }, {}, 1, { [root] = 0 }
    local queue  =  { root, bottom = 1, top = 2 }

    while queue.bottom < queue.top do
        queue.bottom  =  queue.bottom + 1
        local node        =  queue[queue.bottom-1]
        -- Who've been visited? Bookkeeping
        visited[node], info[node]  =  (visited[node] or 0) + 1, info[node] or get_table_info(node)
        if visited[node] == 2 then
            info[node].marker, next_mark  =  next_mark, next_mark + 1
        end
        -- Get table info & visit children.
        if visited[node] < 2 and depth[node] < max_depth then
            for k, v in pairs(node) do
                if type(k) == 'table' then  queue[queue.top], queue.top, depth[k]  =  k, queue.top + 1, math.min(depth[k] or math.huge, depth[node] + 1)  end
                if type(v) == 'table' then  queue[queue.top], queue.top, depth[v]  =  v, queue.top + 1, math.min(depth[v] or math.huge, depth[node] + 1)  end
            end
        end
    end

    -- Use depth collected
    for node in pairs(depth) do
        info[node].depth = depth[node]
        info[node].address = info[node].address .. ': ' .. info[node].depth
    end

    assert(type(info) == 'table')
    return info
end

--------------------------------------------------------------------------------

return { analyze_structure, get_table_info }