feat(scraper): use the new scraper

src/awesomewm.d.tl/init.lua

@@ -3,19 +3,19 @@ local inspect = require "inspect"
 local log = require "logger"
 local properties = require "properties"
 local scraper = require "scraper"
-local generator = require "generator"
+-- local generator = require "generator"
 
 log:info(
    inspect { message = "Start extraction", base_url = properties.base_url }
 )
 
--- local index = crawler.fetch(properties.base_url .. properties.index_uri)
+local index = crawler.fetch(properties.base_url .. properties.index_uri)
--- local modules = scraper.get_modules_from_index(
---    index,
---    properties.ignored_modules
--- )
 
--- log:info(inspect { modules_found = #modules })
+-- local modules =
+--    scraper.get_modules_from_index(index, properties.ignored_modules)
+local module_infos = scraper.module_info_list.get_modules_from_index(index)
+
+log:info(inspect { modules_found = #module_infos })
 
 -- for i = 1, 1 do -- #modules do
 --    local m = modules[i]
@@ -25,12 +25,16 @@ log:info(
 --    log:info(inspect { items })
 -- end
 
-local page =
+local html =
-   crawler.fetch(properties.base_url .. "/widgets/wibox.widget.imagebox.html")
+   crawler.fetch(properties.base_url .. "/widgets/wibox.widget.textbox.html")
-local items = scraper.get_doc_from_page(page)
+local module_doc = scraper.module_doc.get_doc_from_page(html)
+log:info(inspect { module_doc = module_doc })
+
+-- local items = scraper.get_doc_from_page(page)
 -- log:info(inspect { items })
 
-generator.write(
+
-   generator.generate_teal(items),
+-- generator.write(
-   properties.out_directory .. "/test.tl"
+--    generator.generate_teal(items),
-)
+--    properties.out_directory .. "/test.tl"
+-- )

src/awesomewm.d.tl/scraper/init.lua

@@ -1,212 +1,5 @@
-local htmlparser = require "htmlparser"
+return {
-local log = require "logger"
+   module_doc = require "scraper.module_doc",
-local utils = require "utils"
+   module_info_list = require "scraper.module_info_list",
-
+   utils = require "scraper.utils",
-local scraper = {}
+}
-
-function scraper.get_modules_from_index(html, ignored)
-   local document = htmlparser.parse(html)
-
-   local modules = utils.map(document "#navigation ul > li a", function(node)
-      return {
-         name = utils.sanitize_string(node:getcontent()),
-         uri = node.attributes.href,
-      }
-   end)
-
-   local filtered_modules = utils.filter(modules, function(module)
-      return not utils.has_item(ignored, module.name)
-   end)
-
-   return filtered_modules
-end
-
-local function extract_first_node_or_fail(node, selector)
-   local extracted = node(selector)[1]
-
-   if not extracted then
-      log:error {
-         message = "Can't find `" .. selector .. "` element!",
-         node = node:gettext(),
-      }
-      error "extract_first_node_or_fail"
-   end
-
-   return extracted
-end
-
-local function extract_first_node(node, selector)
-   local extracted = node(selector)[1]
-
-   if not extracted then
-      return nil
-   end
-
-   return extracted
-end
-
-local function extract_inner_floating_text(node)
-   local html = node:getcontent()
-
-   -- Remove inner tags from the html
-   for _, n in ipairs(node.nodes) do
-      html = utils.replace(html, n:gettext(), "")
-   end
-
-   return utils.sanitize_string(html)
-end
-
-local function extract_item_content_parameter_list(content)
-   for i, n in ipairs(content.nodes) do
-      -- The parameters <ul> is the next element after one of these <h3>
-      if
-         n.name == "h3"
-         and utils.has_item(
-            { "Parameters:", "Type constraints:", "Arguments" },
-            utils.sanitize_string(n:getcontent())
-         )
-      then
-         return content.nodes[i + 1]
-      end
-   end
-
-   return nil
-end
-
-local function process_item_header_type(node)
-   local types_node = extract_first_node(node, "span.summary_type")
-
-   if not types_node then
-      return nil
-   end
-
-   local types_string = string.match(types_node:getcontent(), "%((.-)%)")
-      or types_node:getcontent()
-   types_string = utils.replace(types_string, "or", ",")
-   local splitted = string.gmatch(types_string, "([^,]+)")
-
-   local types = {}
-   for t in splitted do
-      table.insert(types, utils.sanitize_string(t))
-   end
-
-   return types
-end
-
-local function process_item_header(node)
-   local name_node = extract_first_node_or_fail(node, "strong")
-
-   -- Remove starting ":" character on method name
-   local name = extract_inner_floating_text(name_node):gsub("^:", "")
-   local type = process_item_header_type(node)
-
-   local item_header = {
-      name = name,
-      type = type,
-   }
-
-   return item_header
-end
-
-local function process_item_content_parameters(node)
-   local parameter = {}
-
-   local name = extract_first_node_or_fail(node, "span.parameter")
-   parameter.name = utils.sanitize_string(name:getcontent())
-
-   local nested_parameters_list = extract_first_node(node, "ul")
-   if nested_parameters_list then
-      parameter.type = "table"
-      parameter.nested_parameters = utils.map(
-         nested_parameters_list "* > li",
-         process_item_content_parameters
-      )
-
-      return parameter
-   end
-
-   local types_list = utils.map(node ".type", function(type)
-      return utils.sanitize_string(type:getcontent())
-   end)
-   if #types_list > 0 then
-      parameter.type = types_list
-   end
-
-   return parameter
-end
-
-local function process_item_content(content)
-   local parameters_list = extract_item_content_parameter_list(content)
-   local parameters = parameters_list
-         and utils.map(
-            parameters_list "* > li",
-            process_item_content_parameters
-         )
-      or {}
-
-   local item_content = {
-      parameters = #parameters > 0 and parameters or nil,
-   }
-
-   return item_content
-end
-
-local function process_section_titles(document)
-   return utils.map(document "h2.section-header", extract_inner_floating_text)
-end
-
-local function process_section_items(item)
-   local headers = utils.map(item "dt", process_item_header)
-   local contents = utils.map(item "dd", process_item_content)
-
-   if #headers ~= #contents then
-      log:error {
-         message = "extract_item_content failure: headers and contents don't have the same size",
-         headers = #headers,
-         contents = #contents,
-      }
-      error "extract_item_content"
-   end
-
-   local item_contents = utils.map(headers, function(header, i)
-      return {
-         name = header.name,
-         type = header.type,
-         parameters = contents[i].parameters,
-      }
-   end)
-
-   return item_contents
-end
-
-local function process_section_contents(document)
-   local section_items = document "dl.function"
-
-   local section_contents = utils.map(section_items, process_section_items)
-
-   return section_contents
-end
-
-function scraper.get_doc_from_page(html)
-   local document = htmlparser.parse(html, 9999)
-
-   local sections_titles = process_section_titles(document)
-   local section_contents = process_section_contents(document)
-
-   if #sections_titles ~= #section_contents then
-      log:error {
-         message = "get_items_from_page failure: section_titles and section_contents don't have the same size",
-         sections_titles = #sections_titles,
-         func = #section_contents,
-      }
-      error "get_items_from_page"
-   end
-
-   local doc = utils.map(sections_titles, function(title, i)
-      return { section = title, items = section_contents[i] }
-   end)
-
-   return doc
-end
-
-return scraper