mirror of
https://github.com/koreader/koreader.git
synced 2025-08-10 00:52:38 +00:00
960b2ae62a
By essentially dropping the whole XHTML block, instead of trying to salvage each and every tag one by one as we did before. Also, as that's usually the result after broken parsing, handle nil URLs slightly better in the frontend, so that they get caught/reported properly instead of doing nothing and/or crashing half the time.
100 lines
3.4 KiB
Lua
100 lines
3.4 KiB
Lua
--[[
|
|
This code is derived from the LAPHLibs which can be found here:
|
|
|
|
https://github.com/Wiladams/LAPHLibs
|
|
--]]
|
|
local util = require("util")
|
|
local luxl = require("luxl")
|
|
local ffi = require("ffi")
|
|
|
|
local OPDSParser = {}
|
|
|
|
local unescape_map = {
|
|
["lt"] = "<",
|
|
["gt"] = ">",
|
|
["amp"] = "&",
|
|
["quot"] = '"',
|
|
["apos"] = "'"
|
|
}
|
|
|
|
local gsub = string.gsub
|
|
local function unescape(str)
|
|
return gsub(str, '(&(#?)([%d%a]+);)', function(orig, n, s)
|
|
if unescape_map[s] then
|
|
return unescape_map[s]
|
|
elseif n == "#" then -- unescape unicode
|
|
return util.unicodeCodepointToUtf8(tonumber(s))
|
|
else
|
|
return orig
|
|
end
|
|
end)
|
|
end
|
|
|
|
function OPDSParser:createFlatXTable(xlex, curr_element)
|
|
curr_element = curr_element or {}
|
|
|
|
local curr_attr_name
|
|
local attr_count = 0
|
|
|
|
-- start reading the thing
|
|
for event, offset, size in xlex:Lexemes() do
|
|
local txt = ffi.string(xlex.buf + offset, size)
|
|
if event == luxl.EVENT_START then
|
|
if txt ~= "xml" then
|
|
-- does current element already have something
|
|
-- with this name?
|
|
|
|
-- if it does, if it's a table, add to it
|
|
-- if it doesn't, then add a table
|
|
local tab = self:createFlatXTable(xlex)
|
|
if txt == "entry" or txt == "link" then
|
|
if curr_element[txt] == nil then
|
|
curr_element[txt] = {}
|
|
end
|
|
table.insert(curr_element[txt], tab)
|
|
elseif type(curr_element) == "table" then
|
|
curr_element[txt] = tab
|
|
end
|
|
end
|
|
elseif event == luxl.EVENT_ATTR_NAME then
|
|
curr_attr_name = unescape(txt)
|
|
elseif event == luxl.EVENT_ATTR_VAL then
|
|
curr_element[curr_attr_name] = unescape(txt)
|
|
attr_count = attr_count + 1
|
|
curr_attr_name = nil
|
|
elseif event == luxl.EVENT_TEXT then
|
|
curr_element = unescape(txt)
|
|
elseif event == luxl.EVENT_END then
|
|
return curr_element
|
|
end
|
|
end
|
|
return curr_element
|
|
end
|
|
|
|
function OPDSParser:parse(text)
|
|
-- Murder Calibre's whole "content" block, because luxl doesn't really deal well with various XHTML quirks,
|
|
-- as the list of crappy replacements below attests to...
|
|
-- There's also a high probability of finding orphaned tags or badly nested ones in there, which will screw everything up.
|
|
text = text:gsub('<content type="xhtml">.-</content>', '')
|
|
-- luxl doesn't handle XML comments, so strip them
|
|
text = text:gsub("<!%-%-.-%-%->", "")
|
|
-- luxl prefers <br />, the other two forms are valid in HTML, but will kick luxl's ass
|
|
text = text:gsub("<br>", "<br />")
|
|
text = text:gsub("<br/>", "<br />")
|
|
-- Same deal with hr
|
|
text = text:gsub("<hr>", "<hr />")
|
|
text = text:gsub("<hr/>", "<hr />")
|
|
-- It's also allergic to orphaned <em/> (As opposed to a balanced <em></em> pair)...
|
|
text = text:gsub("<em/>", "")
|
|
-- Let's assume it might also happen to strong...
|
|
text = text:gsub("<strong/>", "")
|
|
-- Some OPDS catalogs wrap text in a CDATA section, remove it as it causes parsing problems
|
|
text = text:gsub("<!%[CDATA%[(.-)%]%]>", function (s)
|
|
return s:gsub( "%p", {["&"] = "&", ["<"] = "<", [">"] = ">" } )
|
|
end )
|
|
local xlex = luxl.new(text, #text)
|
|
return assert(self:createFlatXTable(xlex))
|
|
end
|
|
|
|
return OPDSParser
|