// URLs implements a breadth-first search webcrawler based on the
// example given in section 8.6 of The Go Programming Language.
package main

import (
	"flag"
	"fmt"
	"log"
	"net/url"
	"runtime/pprof"
	"strings"
	"time"

	"git.brandonirizarry.xyz/hamlet"
)

func main() {
	// Setting shorfile helps especially for when we log errors
	// without returning them.
	log.SetFlags(log.LstdFlags | log.Lshortfile)

	maxConcurrency := flag.Int("c", 0, "Maximum number of concurrent queue pushes")
	urlArg := flag.String("url", "", "Entry-point URL")
	maxURLs := flag.Int("max", 0, "Maximum number of URLs to collect (omitted or 0 means no limit)")
	maxDepth := flag.Int("depth", 0, "Maximum URL depth (omitted or 0 means no limit)")
	shortcode := flag.String("shortcode", "", "URL shortcode")
	shortcodeFilename := flag.String("scfile", "urls.csv", "Shortcode CSV file")

	flag.Parse()

	// Vet the given CLI arguments for things like negative or
	// missing values.
	if *maxConcurrency == 0 {
		log.Fatal("Missing -c argument")
	}

	if *maxConcurrency < 1 {
		log.Fatalf("Invalid -c argument: %d", *maxConcurrency)
	}

	if *maxURLs < 0 {
		log.Fatalf("Invalid -max argument: %d", *maxURLs)
	}

	if *maxDepth < 0 {
		log.Fatalf("Invalid -depth argument: %d", *maxDepth)
	}

	// Use hamlet to vet the validity of the given command-line
	// argument configuration provided by the user.
	shortcodePresent := hamlet.Present(*shortcode)
	urlArgPresent := hamlet.Present(*urlArg)
	if hamlet.Xor(urlArgPresent, shortcodePresent).Deny() {
		log.Fatal("Need exactly one of -url or -shortcode")
	}

	if hamlet.If(hamlet.Present(*shortcodeFilename), shortcodePresent).Deny() {
		log.Fatal("-scfile implies use of -shortcode")
	}

	startRawURL := *urlArg

	if shortcodePresent.Assert() {
		var err error
		startRawURL, err = getURLFromShortcode(*shortcodeFilename, *shortcode)
		if err != nil {
			log.Fatal(err)
		}
	}

	startURL, err := convertToURL(startRawURL)
	if err != nil {
		log.Fatal(err)
	}

	// Our web crawlers use concurrency: check if any goroutines
	// have leaked.
	getLeakProfile(func() {
		classic(startURL, *maxConcurrency, *maxURLs, *maxDepth)
	})
}

// convertToURL parses the given rawURL into a [url.URL]. If the
// rawURL is missing a scheme, "https://" is prepended before parsing.
//
// Return the parsed URL, along with any error.
func convertToURL(rawURL string) (url.URL, error) {
	if !strings.HasPrefix(rawURL, "http://") && !strings.HasPrefix(rawURL, "https://") {
		rawURL = "https://" + rawURL
		fmt.Printf("start url: %s\n", rawURL)
	}

	u, err := url.Parse(rawURL)
	if err != nil {
		return url.URL{}, fmt.Errorf("can't parse %s: %w", rawURL, err)
	}

	return *u, nil
}

// getLeakProfile runs a leaky program snippet, extracts the goroutine leak profile,
// and writes it to stdout.
func getLeakProfile(leakySnippet func()) {
	prof := pprof.Lookup("goroutineleak")
	defer func() {
		time.Sleep(2 * time.Second)
		var content strings.Builder

		prof.WriteTo(&content, 2)
		// Ignore non leaked goroutines
		leaks := strings.SplitSeq(content.String(), "\n\n")
		for leak := range leaks {
			if strings.Contains(leak, "(leaked)") {
				fmt.Println(leak + "\n")
			}
		}
	}()

	leakySnippet()
}