path: root/c/data.cpp

#include <thread>
#include <windows.h>
#include <wininet.h>
#include <libxml/HTMLparser.h>
#include <libxml/HTMLtree.h>
#include <libxml/xpath.h>

#include "data.h"
#include "episodelistview.h"
#include "err.h"
#include "res.h"
#include "util.h"
#include "win32.h"
#include "window.h"

using namespace std::string_literals;

UniqueOk<htmlParserCtxtPtr, xmlFreeParserCtxt> RemoteParserCtxt(const wchar_t* wszUrl, const char* szUrl)
{
	static Unique<HINTERNET, InternetCloseHandle> hi =
		InternetOpenW(L"Episode Browser", INTERNET_OPEN_TYPE_DIRECT, nullptr, nullptr, 0);
	if (hi.Bad(0))
		throw Err(WINDOWS, L"Internet handle could not be opened");

	Unique<HINTERNET, InternetCloseHandle> hiUrl = InternetOpenUrlW(
	    hi.v, wszUrl, nullptr, 0, INTERNET_FLAG_NO_UI, 0);
	if (hiUrl.Bad(0))
		throw Err(WININET, L"Could not open "s + wszUrl + L"");

	char bufX[1024];
	Unique<htmlParserCtxtPtr, xmlFreeParserCtxt> ctx = htmlCreatePushParserCtxt(
	    nullptr, nullptr, bufX, sizeof(bufX), szUrl, XML_CHAR_ENCODING_UTF8);
	if (ctx.Bad(0))
		throw Err(LIBXML2, L"HTML parser context could not be created");

	htmlCtxtUseOptions(ctx.v, HTML_PARSE_RECOVER|HTML_PARSE_NOERROR|HTML_PARSE_NOWARNING);

	BOOL r;
	DWORD cbRead;
	char bufI[1024];
	while (r = InternetReadFile(hiUrl.v, bufI, sizeof(bufI), &cbRead), cbRead) {
		if (!r)
			throw Err(WININET, L"HTML could not be retrieved");
		if (!htmlParseChunk(ctx.v, bufI, cbRead, 0))
			throw Err(LIBXML2, L"HTML could not be parsed");
	}
	htmlParseChunk(ctx.v, bufI, 0, 1); /* Stop parsing. */

	return ctx;
}

static inline void XmlFree(void* p) { xmlFree(p); }

template <size_t N>
bool WcharsFromXmlchars(wchar_t (&dst)[N], Unique<xmlChar*, XmlFree> utf8)
{
	if (utf8.Bad(0))
		throw Err(LIBXML2, L"Node content could not be retrieved");

	/* Truncate if source is larger than destination. */
	int lenUtf8 = xmlStrlen(utf8.v);
	utf8.v[Min(N, static_cast<size_t>(lenUtf8))] = 0;

	/* Convert internal representation from UTF-8 to Latin-1,
	 * which seems to actually convert the string to proper UTF-8
	 * (???). */
	unsigned char lat1[N];
	int lenLat1 = N-1;
	if (UTF8Toisolat1(lat1, &lenLat1, utf8.v, &lenUtf8) <= 0)
		return false;
	lat1[lenLat1] = 0;

	/* Write wide string to destination, if it fits. */
	char* const src = reinterpret_cast<char*>(lat1);
	const int cchNarrow = lenLat1+1;
	const int cchWide = MultiByteToWideChar(CP_UTF8, 0, src, cchNarrow, nullptr, 0);
	if (static_cast<size_t>(cchWide) > N)
		return false;
	return MultiByteToWideChar(CP_UTF8, 0, src, cchNarrow, dst, cchWide);
}

/* The Fetch* functions are run in a separate thread via WaitFor. The
 * main thread and the fetch thread communicate by setting flags on a
 * shared byte. At any given time, only a single fetch thread may be
 * performing work. */

enum Signal : unsigned char
{
	READY = 1<<0,	/* Main -> fetch: start working! */
	DONE = 1<<1,	/* Fetch -> main: work is done. */
	ABORT = 1<<2	/* Main -> fetch: exit prematurely! */
};

static Window* s_window;

void WaitFor(Window& window, void (*f)(unsigned char*))
{
	static unsigned char sig = READY;
	static UINT_PTR iTimer = 0;

	static auto procTimer = [](HWND, UINT, UINT_PTR, DWORD) -> void
	{
		static int i = 0;

		if (sig & DONE) {
			KillTimer(nullptr, iTimer);
			i = 0;
			sig = READY; /* Reset signals. */
			s_window->elv.Update(); /* Reset status bar. */
			EnableMenuItem(GetMenu(s_window->hWnd), IDM_FILE_FETCH_CANCEL, MF_GRAYED);
		} else {
			/* Animate ellipsis in status bar. */
			static const wchar_t* text[] = {L".", L"..", L"...", L""};
			i = (i+1)%(sizeof(text)/sizeof(*text));
			s_window->Status(text[i], 1);
		}
	};

	static auto procThread = [](void (*f)(unsigned char*)) noexcept -> void
	{
		std::set_terminate(OnTerminate);
		while (!(sig & READY))
			Sleep(100);
		sig = 0;
		EnableMenuItem(GetMenu(s_window->hWnd), IDM_FILE_FETCH_CANCEL, MF_ENABLED);
		try {
			f(&sig);
			sig |= DONE;
		} catch (...) {
			sig |= DONE;
			EBMessageBox(What(), L"Remote Data Retrieval Error");
		}
	};

	/* Null indicates that any active task should be cancelled. */
	if (!f) {
		sig |= ABORT;
		EnableMenuItem(GetMenu(s_window->hWnd), IDM_FILE_FETCH_CANCEL, MF_GRAYED);
		return;
	}

	/* Ensure that only a single thread is waited on. */
	if (!(sig & READY)) {
		if (EBMessageBox(L"Another task is active. "
		    L"Do you want to cancel the existing task and start a new one?",
		    L"Error", MB_YESNO|MB_ICONWARNING) == IDYES)
			sig |= ABORT;
		else
			return;
	}

	s_window = &window;
	std::thread(procThread, f).detach();
	s_window->Status(L".", 1);
	if (!(iTimer = SetTimer(nullptr, iTimer, 500, procTimer)))
		throw Err(WINDOWS, L"Timer could not be started");
}

void FetchData(unsigned char* sig)
{
	/* The remote data is retrieved using WinINet from the
	 * Detective Conan World wiki. Using libxml2's "push parser",
	 * the HTML is parsed piece by piece as it is retrieved. The
	 * episode data are contained in table rows matching a (very!)
	 * specific XPath query. This is fragile theoretically, but
	 * unlikely to break practically. */

	wchar_t url[_countof(s_window->cfg.prefixUrl)+46+1];
	Swprintf(url, L"%shttps://www.detectiveconanworld.com/wiki/Anime", s_window->cfg.prefixUrl);

	UniqueOk<htmlParserCtxtPtr, xmlFreeParserCtxt> ctx =
		RemoteParserCtxt(url, "https://www.detectiveconanworld.com/wiki/Anime");

	Unique<xmlXPathContextPtr, xmlXPathFreeContext> xpathCtx = xmlXPathNewContext(ctx.v->myDoc);
	if (xpathCtx.Bad(0))
		throw Err(LIBXML2, L"XPath context could not be created");

	Unique<xmlXPathObjectPtr, xmlXPathFreeObject> xpathObj = xmlXPathEvalExpression(
	    reinterpret_cast<const xmlChar*>("//tr[./td[1] != '' and ./td[3][@style='background:#f2fde9;']]"),
	    xpathCtx.v);
	if (xpathObj.Bad(0))
		throw Err(LIBXML2, L"XPath object could not be created");

	xmlNodeSetPtr nodes = xpathObj.v->nodesetval;

	if (!nodes || !nodes->nodeNr)
		throw Err(GENERIC, L"Data retrieval failed: No matching HTML nodes found");

	for (int i = 0; i < nodes->nodeNr; i++) {
		if (*sig & ABORT)
			return;

		const xmlNodePtr node = nodes->nodeTab[i];
		if (xmlChildElementCount(node) != 8)
			throw Err(GENERIC, L"Data retrieval failed: Unexcepted number of columns in table");

		ElvDataA& e = s_window->fvElv.At(i);
		DlvDataA& d = s_window->fvDlv.At(i);

		/* Each datum is contained within a specific cell in
		 * the row. The child element count above ensures that
		 * none of the following nodes are null. */
		const xmlNodePtr nodeEp = xmlFirstElementChild(node);
		const xmlNodePtr nodeTitle = xmlNextElementSibling(xmlNextElementSibling(nodeEp));
		const xmlNodePtr nodeDate = xmlNextElementSibling(nodeTitle);
		const xmlNodePtr nodeSource = xmlNextElementSibling(
		    xmlNextElementSibling(xmlNextElementSibling(nodeDate)));
		const xmlNodePtr nodeHint = xmlNextElementSibling(nodeSource);

		WcharsFromXmlchars(d.date, xmlNodeGetContent(nodeDate));
		WcharsFromXmlchars(d.source, xmlNodeGetContent(nodeSource));
		WcharsFromXmlchars(d.hint, xmlNodeGetContent(nodeHint));
		e.bTVOriginal = wcsncmp(d.source, L"TV", 2) == 0? 1: 0;
		WcharsFromXmlchars(e.siEp, xmlNodeGetContent(nodeEp));
		e.siEp[wcscspn(e.siEp, L"W")] = 0; /* Remove potential "WPS" suffix. */
		WcharsFromXmlchars(e.title, xmlNodeGetContent(nodeTitle));

		/* Retrieve the link to the episode's wiki entry,
		 * which should be the first (and only) child element
		 * of the title node. */
		const xmlNodePtr nodeLink = xmlFirstElementChild(nodeTitle);
		if (nodeLink)
			WcharsFromXmlchars(d.wiki,
			    xmlGetProp(nodeLink, reinterpret_cast<const xmlChar*>("href")));
	}
}

void FetchScreenwriters(unsigned char* sig)
{
	/* Screenwriters are expensive to fetch, so we try to avoid
	 * fetching screenwriters for episodes that already have a
	 * screenwriter. Additionally, in the same session, we don't
	 * try to fetch screenwriters for episodes for which we have
	 * already tried to fetch screenwriters. We keep track of
	 * these states using the iLast variable. */
	static int iLast = -1;
	int iMax = s_window->cfg.cEp-1;

	/* Find the last episode that has a screenwriter. */
	if (iLast == -1)
		for (size_t i = 0; i < s_window->fvDlv.c; i++)
			if (const DlvDataA& d = s_window->fvDlv[i]; !d.date[0]) {
				iMax = i-1;
				break;
			} else if (d.screenwriter[0])
				iLast = i;

	FINALLY { s_window->Status(L""); };

	/* Fetch screenwriters for the rest of the episodes. */
	const wchar_t prefix[] = L"https://www.detectiveconanworld.com";
	wchar_t url[256];
	Wcscpy(url, prefix);

	Unique<xmlXPathContextPtr, xmlXPathFreeContext> xpathCtx;
	Unique<xmlXPathObjectPtr, xmlXPathFreeObject> xpathObj;
	for (iLast++; iLast < iMax; iLast++) {
		if (*sig & ABORT)
			return;

		wchar_t msg[48];
		Swprintf(msg, L"Fetching screenwriter for episode %d...", iLast+1);
		s_window->Status(msg);

		/* Retrieve URL for episode's wiki page. */
		DlvDataA& d = s_window->fvDlv[iLast];
		Wcscpy(Buf(url)+Len(prefix), d.wiki);

		/* Retrieve screenwriter from HTML. */
		UniqueOk<htmlParserCtxtPtr, xmlFreeParserCtxt> ctx = RemoteParserCtxt(url, nullptr);
		xpathCtx = xmlXPathNewContext(ctx.v->myDoc);
		if (xpathCtx.Bad(0))
			throw Err(LIBXML2, L"XPath context could not be created");

		xpathObj = xmlXPathEvalExpression(reinterpret_cast<const xmlChar*>(
			    "//th[contains(text(), 'Screenplay:')]/following-sibling::td"),
			    xpathCtx.v);
		if (xpathObj.Bad(0))
			throw Err(LIBXML2, L"XPath object could not be created");

		xmlNodeSetPtr nodes = xpathObj.v->nodesetval;
		if (nodes && nodes->nodeNr)
			WcharsFromXmlchars(d.screenwriter, xmlNodeGetContent(nodes->nodeTab[0]));
	}
}