Changeset 33965

Timestamp:

2020-02-21T20:59:07+13:00 (4 years ago)

Author:

ak19

Message:

1. Adding a basicDomain column (stripped of http/https and www prefix) for easier sorting in LibreOffice Calc spreadshoot. 2. Adding a recordCount column to maintain a way to get back the original order after resorting in LibreOffice Calc.

File:

• other-projects/maori-lang-detection/src/org/greenstone/atea/ManualURLInspection.java (modified) (4 diffs)

other-projects/maori-lang-detection/src/org/greenstone/atea/ManualURLInspection.java

@@ -641 +641 @@
          CSVPrinter csvWriter = new CSVPrinter(new FileWriter(tmpOutFile), CSVFormat.DEFAULT.withQuoteMode(QuoteMode.MINIMAL));
          ) {
-
+        
+        
         int recordCount = 0;
+
+        csvWriter.printRecord("origSequence", "basicDomain","pageURL",
+                  "countryCode","mostlyMRI","qualityLevel",
+                  "numPagesInMRIForDomainSuffix","totalPagesInMRIForDomainSuffix");
+
+        
         for (CSVRecord csvRecord : parser) {        
         
@@ -649 +656 @@
             continue;
         }
-        
         recordCount++;
         
         String basicURL = Utility.stripProtocolAndWWWFromURL(Utility.getDomainForURL(url, false));
-
+        
         String countryCode = csvRecord.get(COUNTRY_CODE_COLUMN);
         String isReallyInMRI = "";
@@ -677 +683 @@
 
         // Save the CSV record into the tmp file with the 2 counts columns
-        csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel,
+        csvWriter.printRecord(recordCount, basicURL,
+                      url,  countryCode, isReallyInMRI, qualityLevel,
                       countNumPagesInMRI, countTotalPages);     
         }
@@ -694 +701 @@
     System.err.println("Usage: ManualURLInspection webPageURLs.txt");
     }
-
+    
+    //^https?://(www.)?
     /**
      * If no args are passed in, generates complete containsMRI file listings for NZ and overseas web SITES (domains),