protected async override Task <ProcessResult> OnPolling(PollerJobParameters parameters, string workingFolder)
{
//if this handle already has rasterImage we do not need to do anything
//we can skip only if this action is not forced or if the pdf was not generated by office
//because office file can be re-generated due to errors in conversoin (libreoffice or MsOffice)
if (!IsForced(parameters) && !FromPipelineId(parameters, "office"))
{
var formats = GetFormats(parameters.TenantId, parameters.JobId);
if (formats.Any(f => f == DocumentFormats.RasterImage))
{
return(ProcessResult.Ok);
}
}
String format = parameters.All[JobKeys.ThumbnailFormat];
Logger.DebugFormat("Conversion for jobId {0} in format {1} starting", parameters.JobId, format);
var task = _taskFactory();
var passwords = ClientPasswordSet.GetPasswordFor(parameters.FileName);
foreach (var password in passwords)
{
task.Passwords.Add(password);
}
string pathToFile = await DownloadBlob(parameters.TenantId, parameters.JobId, parameters.FileName, workingFolder).ConfigureAwait(false);
var convertParams = new CreatePdfImageTaskParams()
{
Dpi = parameters.GetIntOrDefault(JobKeys.Dpi, 150),
FromPage = parameters.GetIntOrDefault(JobKeys.PagesFrom, 1),
Pages = parameters.GetIntOrDefault(JobKeys.PagesCount, 1),
Format = (CreatePdfImageTaskParams.ImageFormat)Enum.Parse(typeof(CreatePdfImageTaskParams.ImageFormat), format, true)
};
await task.Run(
pathToFile,
convertParams,
(i, s) => Write(workingFolder, parameters, format, i, s) //Currying
).ConfigureAwait(false);
Logger.DebugFormat("Conversion of {0} in format {1} done", parameters.JobId, format);
return(ProcessResult.Ok);
}
protected async override Task <ProcessResult> OnPolling(
PollerJobParameters parameters,
String workingFolder)
{
Boolean result;
var contentFileName = Path.ChangeExtension(parameters.FileName, ".content");
if (!_formats.Contains(parameters.FileExtension))
{
Logger.DebugFormat("Document for job Id {0} has an extension not supported, setting null content", parameters.JobId);
return(new ProcessResult(await AddNullContentFormat(parameters, contentFileName)));
}
Logger.DebugFormat("Starting tika on job: {0}, file extension {1}", parameters.JobId, parameters.FileExtension);
Logger.DebugFormat("Downloading blob for job: {0}, on local path {1}", parameters.JobId, workingFolder);
string pathToFile = await DownloadBlob(parameters.TenantId, parameters.JobId, parameters.FileName, workingFolder);
pathToFile = ProcessFile(pathToFile, workingFolder);
Boolean shouldAnalyze = _filterManager.ShouldAnalyze(parameters.FileName, pathToFile);
if (!shouldAnalyze)
{
Logger.InfoFormat("File {0} for job {1} was discharded!", parameters.FileName, parameters.JobId);
return(new ProcessResult(await AddNullContentFormat(parameters, contentFileName)));
}
Logger.DebugFormat("Search for password JobId:{0}", parameters.JobId);
var passwords = ClientPasswordSet.GetPasswordFor(parameters.FileName).ToArray();
String content = "";
Int32 analyzerOrdinal = 0;
Boolean success = false;
var analyzer = BuildAnalyzer(analyzerOrdinal);
do
{
try
{
if (passwords.Any())
{
//Try with all the password
foreach (var password in passwords)
{
try
{
content = analyzer.GetHtmlContent(pathToFile, password) ?? "";
break; //first password that can decrypt file break the list of password to try
}
catch (Exception)
{
Logger.ErrorFormat("Error opening file {0} with password", parameters.FileName);
}
}
}
else
{
//Simply analyze file without password
Logger.DebugFormat("Analyze content JobId: {0} -> Path: {1}", parameters.JobId, pathToFile);
content = analyzer.GetHtmlContent(pathToFile, "") ?? "";
}
success = true;
}
catch (Exception ex)
{
Logger.ErrorFormat(ex, "Error extracting tika with analyzer {0} on file {1}", analyzer.Describe(), parameters.FileName, parameters.JobId);
analyzer = BuildAnalyzer(++analyzerOrdinal);
if (analyzer != null)
{
Logger.InfoFormat("Retry job {0} with analyzer {1}", parameters.JobId, analyzer.Describe());
}
}
} while (analyzer != null && success == false);
Logger.DebugFormat("Finished tika on job: {0}, charsNum {1}", parameters.JobId, content.Count());
String sanitizedContent = content;
if (!string.IsNullOrWhiteSpace(content))
{
var resultContent = _builder.CreateFromTikaPlain(content);
var documentContent = resultContent.Content;
sanitizedContent = resultContent.SanitizedTikaContent;
var pages = documentContent.Pages.Count();
string lang = null;
if (pages > 1)
{
lang = LanguageDetector.GetLanguage(documentContent.Pages[1].Content);
}
if (lang == null && pages == 1)
{
lang = LanguageDetector.GetLanguage(documentContent.Pages[0].Content);
}
if (lang != null)
{
documentContent.AddMetadata(DocumentContent.MedatataLanguage, lang);
}
result = await AddFormatToDocumentFromObject(
parameters.TenantId,
this.QueueName,
parameters.JobId,
new DocumentFormat(DocumentFormats.Content),
documentContent,
contentFileName,
new Dictionary <string, object>());
Logger.DebugFormat("Added format {0} to jobId {1}, result: {2}", DocumentFormats.Content, parameters.JobId, result);
}
var tikaFileName = Path.Combine(workingFolder, Path.GetFileNameWithoutExtension(parameters.FileName) + ".tika.html");
tikaFileName = SanitizeFileNameForLength(tikaFileName);
File.WriteAllText(tikaFileName, sanitizedContent);
result = await AddFormatToDocumentFromFile(
parameters.TenantId,
parameters.JobId,
new DocumentFormat(DocumentFormats.Tika),
tikaFileName,
new Dictionary <string, object>());
Logger.DebugFormat("Added format {0} to jobId {1}, result: {2}", DocumentFormats.Tika, parameters.JobId, result);
return(ProcessResult.Ok);
}
