public async Task <Incident> LoadAsync()
{
try
{
var handler = new HttpClientHandler
{
AutomaticDecompression = DecompressionMethods.Deflate | DecompressionMethods.GZip
};
var client = new HttpClient(handler);
// setting the default user agent
if (client.DefaultRequestHeaders.UserAgent.Count == 0)
{
client.DefaultRequestHeaders.UserAgent.ParseAdd("Azure Function");
}
var httpResponseMessage = await client.GetAsync(_uri);
httpResponseMessage.EnsureSuccessStatusCode();
// There is a bug in the .net framework that causes ReadAsStringAsync() to fail if the server reports the content encoding as "utf-8" rather than utf-8 https://github.com/dotnet/corefx/issues/5014
if (httpResponseMessage.Content.Headers.ContentType?.CharSet == @"""utf-8""")
{
httpResponseMessage.Content.Headers.ContentType.CharSet = "UTF-8";
}
var data = await httpResponseMessage.Content.ReadAsStringAsync();
data = StringSanitizer.SimplifyHtmlEncoded(data);
var sr = new SmartReader.Reader(_uri, data);
sr.AddCustomOperationStart(SpaceElements);
var article = sr.GetArticle();
var content = !string.IsNullOrEmpty(article.TextContent) ? article.TextContent : article.Excerpt;
if (!string.IsNullOrEmpty(content))
{
var shortSummary =
StringSanitizer.RemoveDoublespaces(
StringSanitizer.RemoveUrls(
StringSanitizer.RemoveHashtags(content)))
.Trim();
var summary =
StringSanitizer.RemoveDoublespaces(content)
.Trim();
return(new Incident(shortSummary, summary));
}
}
catch (Exception e)
{
_logger.LogError(e, "Exception loading article");
}
return(null);
}
// Clean up for the ShortSummary
// The short summary is the shortened version of the summary that is optimized to be processed by LUIS
private string ShortSummaryCleanUp(string body)
{
string result = StringSanitizer.RemoveHtmlTags(body);
result = StringSanitizer.SimplifyHtmlEncoded(result);
result = StringSanitizer.RemoveFillerWords(result);
result = StringSanitizer.RemoveSpecialCharacters(result);
result = StringSanitizer.RemoveDoublespaces(result);
return(result.Trim());
}
// Clean up for the summary
// The summary is the human readable content that is displayed in the ESRI portal
private string SummaryCleanUp(string body)
{
string result = StringSanitizer.RemoveHtmlTags(body);
result = WebUtility.HtmlDecode(result);
result = StringSanitizer.RemoveHashtags(result);
result = StringSanitizer.RemoveDoublespaces(result);
result = result.Trim();
return(result);
}
public static async Task <LuisInput> Run([QueueTrigger("twitter")] string json, ILogger log)
{
log.LogInformation($"Scrape function invoked:\n{json}");
var tweet = JsonConvert.DeserializeObject <TweetModel>(json);
string summary = null, shortSummary = null;
var sourceUrl = await SourceUrlFromTweetAsync(tweet, log);
if (!string.IsNullOrEmpty(sourceUrl))
{
log.LogInformation($"Loading external reference into scraper '{sourceUrl}'.");
var incident = await GetIncidentFromUrl(log, sourceUrl);
if (incident != null)
{
shortSummary = incident.ShortSummary;
summary = incident.Summary;
}
else
{
log.LogWarning($"Reference failed to load content from '{sourceUrl}'.");
}
}
else
{
log.LogInformation("No source url was available for this input, skipping scrape.");
string text = StringSanitizer.SimplifyPunctuation(
System.Net.WebUtility.HtmlDecode(tweet.TweetText));
summary = StringSanitizer.RemoveDoublespaces(text);
shortSummary =
StringSanitizer.RemoveDoublespaces(
StringSanitizer.RemoveUrls(
StringSanitizer.RemoveHashtags(text)));
}
return(new LuisInput
{
SourceUrl = sourceUrl,
TwitterProfileUrl = $"https://twitter.com/{tweet.TweetedBy}",
TweetUrl = $"https://twitter.com/{tweet.TweetedBy}/status/{tweet.TweetId}",
UserLocation = tweet.UserDetails?.Location,
ShortSummary = shortSummary,
Summary = summary
});
}
