Accumulo is a big key/value store. Writing data to Accumulo is flexible and fast. Like any database, Accumulo stores @@ -227,7 +227,7 @@ if (e.keyCode == '39') { window.location = '/tour/data-model/'; } < - 2 / 9 + 2 / 11 > diff --git a/tour/batch-scanner-code/index.html b/tour/batch-scanner-code/index.html index 11b0680..6e8559d 100644 --- a/tour/batch-scanner-code/index.html +++ b/tour/batch-scanner-code/index.html @@ -153,7 +153,7 @@
Ranges and Splits

Conditional Writer Code

Accumulo Tour: Ranges and Splits

Tour page 7 of 9
Accumulo Tour: Conditional Writer Code

Tour page 11 of 11
A Range is a specified group of Keys. There are many different ways to create a Range. Here are a few examples:
Below is a solution to the excercise.
+ +
  static boolean setAddress(Connector conn, String id, String expectedAddr, String try (ConditionalWriter writer = conn.createConditionalWriter("GothamPD", new ConditionalWriterConfig())) {
+      Condition condition = new Condition("location", "home");
+      if(expectedAddr != null) {
+        condition.setValue(expectedAddr);
+      }
+      ConditionalMutation mutation = new ConditionalMutation(id, condition);
+      mutation.put("location", "home", newAddr);
+      return writer.write(mutation).getStatus() == Status.ACCEPTED;
+    } catch (Exception e) {
+      throw new RuntimeException(e);
+    }
+  }
 
-Range r1 = new Range(startKey, endKey);  // Creates a range from startKey inclusive to endKey inclusive.
-Range r2 = new Range(row);               // Creates a range that covers an entire row.
-Range r3 = new Range(startRow, endRow);  // Creates a range from startRow inclusive to endRow inclusive.
 
 
 
-A Scanner by default will scan all Keys in a table but this can be inefficient. It is a good practice to set a range on a Scanner.
-
-scanner.setRange(new Range("id0000", "id0010"));  // returns rows from id0000 to id0010
+The following output shows running the example with a conditional writer.
+Threads retry when conditional mutations are rejected.  The final address has
+all three modifications.
+
+Thread  37 attempting change '  1007 Mountain Drive, Gotham, New York  ' -> '  1007 Mountain Dr, Gotham, New York  '
+Thread  38 attempting change '  1007 Mountain Drive, Gotham, New York  ' -> '1007 Mountain Drive, Gotham, New York'
+Thread  39 attempting change '  1007 Mountain Drive, Gotham, New York  ' -> '  1007 Mountain Drive, Gotham, NY  '
+Thread  38 attempting change '  1007 Mountain Dr, Gotham, New York  ' -> '1007 Mountain Dr, Gotham, New York'
+Thread  39 attempting change '  1007 Mountain Dr, Gotham, New York  ' -> '  1007 Mountain Dr, Gotham, NY  '
+Thread  39 attempting change '1007 Mountain Dr, Gotham, New York' -> '1007 Mountain Dr, Gotham, NY'
+Final address : '1007 Mountain Dr, Gotham, NY'
 
 
 
-As your data grows larger, Accumulo will split tables into smaller pieces called Tablets.  Tablets can then be distributed across multiple Tablet Servers.

-By default, a table will get split into Tablets on row boundaries, guaranteeing an entire row to be on one Tablet Server.  We have the ability to 
-tell Accumulo where to split tables by setting split points. This is done using addSplits in the TableOperations API.  The image below 
-demonstrates how Accumulo splits data.
-
-
-
-Take a minute to learn these Accumulo terms:
-
-  Tablet - A partition of a table.
-  Split - A point where tables are partitioned into separate tablets.
-  Flush - Action taken when data is written from memory to disk.
-  Compact - Action taken when files on disk are consolidated.
-  Iterator - A server side mechanism that can filter and modify Key/Value pairs.
-
-
-Knowing these terms are critical when working closely with Accumulo.  Iterators are especially unique and powerful.  More on them later.
-
-When working with large amounts of data across many Tablet Servers, a simple Scanner might not do the trick. Next lesson we learn about the power of 
-the multi-threaded BatchScanner!
-
-
 

 
 
 
 
   
     
-    <
+    <
     
 
-    7 / 9
-    
-    >
+    11 / 11
     
   
 
diff --git a/tour/conditional-writer/index.html b/tour/conditional-writer/index.html
index 7b53dcb..a464a35 100644
--- a/tour/conditional-writer/index.html
+++ b/tour/conditional-writer/index.html
@@ -153,26 +153,156 @@
 
 
   Accumulo Tour: Conditional Writer
-  Tour page  of 9
+  Tour page 10 of 11
 
 
-  
+  Suppose the Gotham PD is storing home addresses for persons of interest in
+Accumulo.  We want to correctly handle the case of multiple users editing the
+same address at the same time. The following sequence of events shows an example
+of how this can go wrong.
+
+
+  User 0 sets the key id0001:location:home to 1007 Mountain Drive, Gotham, New York
+  User 1 reads id0001:location:home
+  User 2 reads id0001:location:home
+  User 1 replaces Drive with Dr
+  User 2 replaces New York with NY
+  User 1 sets key id0001:location:home to 1007 Mountain Dr, Gotham, New York
+  User 2 sets key id0001:location:home to 1007 Mountain Drive, Gotham, NY
+
+
+In this situation the changes made by User 1 are lost, ending up with 1007
+Mountain Drive, Gotham, NY instead of 1007 Mountain Dr, Gotham, NY.  To
+correctly handle this, Accumulo offers the ConditionalWriter.  The
+ConditionalWriter atomically checks conditions on a row and only applies a
+mutation when all conditions are satisfied.
+
+Exercise
+
+The following code simulates the concurrency in the situation above.  The code
+starts multiple threads, with each thread doing the following.
+
+
+  Read key’s value into memory using a scanner
+  Modify the copy in memory.
+  Write out the modified value from memory using a batch writer.
+  If write was unsuccessful, then goto step 1.
+
+
+This process can result in threads overwriting each other changes.  The problem
+is the batch writer always makes the update, even when the value has
+changed since it was read.
+
+  static String getAddress(Connector conn, String id) {
+    // The IsolatedScanner ensures partial changes to a row are not seen
+    try (Scanner scanner = new IsolatedScanner(conn.createScanner("GothamPD", Authorizations.EMPTY< [...]
+      scanner.setRange(Range.exact(id, "location", "home"));
+      for (Entry<Key,Value> entry : scanner) {
+        return entry.getValue().toString();
+      }
+      return null;
+    } catch (TableNotFoundException e) {
+      throw new RuntimeException(e);
+    }
+  }
+
+  static boolean setAddress(Connector conn, String id, String expectedAddr, String newAddr) {
+    try (BatchWriter writer = conn.createBatchWriter("GothamPD", new BatchWriterConfig())) {
+      Mutation mutation = new Mutation(id);
+      mutation.put("location", "home", newAddr);
+      writer.addMutation(mutation);
+      return true;
+    } catch (Exception e) {
+      throw new RuntimeException(e);
+    }
+  }
+
+  public static Future<Void> modifyAddress(Connector conn, String id, Function<Stringreturn CompletableFuture.runAsync(() -> {
+      String currAddr, newAddr;
+      do {
+        currAddr = getAddress(conn, id);
+        newAddr = modifier.apply(currAddr);
+        System.out.printf("Thread %3d attempting change %20s -> %-20s\n",
+            Thread.currentThread().getId(), "'"+currAddr+"'", "'"+newAddr+"'");
+      } while (!setAddress(conn, id, currAddr, newAddr));
+    });
+  }
+
+  static void exercise(MiniAccumuloCluster mac) throws Exception {
+    Connector conn = mac.getConnector("root", "tourguide");
+    conn.tableOperations().create("GothamPD");
+
+    String id = "id0001";
+
+    setAddress(conn, id, null, "  1007 Mountain Drive, Gotham, New York  ");
+
+    // create async operation to trim whitespace
+    Future<Void> future1 = modifyAddress(conn, id, String::trim);
+
+    // create async operation to replace Dr with Drive
+    Future<Void> future2 = modifyAddress(conn, id, addr -> addr.replace(// create async operation to replace New York with NY
+    Future<Void> future3 = modifyAddress(conn, id, addr -> addr.replace(// wait for async operations to complete
+    future1.get();
+    future2.get();
+    future3.get();
+
+    // print the address stored in Accumulo
+    System.out.println("Final address : '"+getAddress(conn, id)+"'");
+  }
+
+
+
+The following is one of a few possible outputs.  Notice that only the
+modification of Drive to Dr shows up in the final output.  The other
+modifications were lost.
+
+Thread  36 attempting change '  1007 Mountain Drive, Gotham, New York  ' -> '1007 Mountain Drive, Gotham, New York'
+Thread  38 attempting change '  1007 Mountain Drive, Gotham, New York  ' -> '  1007 Mountain Drive, Gotham, NY  '
+Thread  37 attempting change '  1007 Mountain Drive, Gotham, New York  ' -> '  1007 Mountain Dr, Gotham, New York  '
+Final address : '  1007 Mountain Dr, Gotham, New York  '
+
+
+
+To fix this example, make the following changes in setAddress() to use a
+ConditionalWriter.
+
+
+  Call createConditionalWriter instead of creating a batch writer
+  Create a Condition for the column ‘location:home’.  If expectedAddr is not null, then call setValue passing expectedAddr.  If [...]
+  
Replace Mutation with a ConditionalMutation and pass the condition to its constructor.
+  Call write passing it the conditional mutation.
+  Return true if getStatus from the Result returned by 
   
     
+    <
+    
 
-     / 9
+    10 / 11
+    
+    >
     
   
 
diff --git a/tour/data-model-code/index.html b/tour/data-model-code/index.html
index 1fd5f6b..1527c6d 100644
--- a/tour/data-model-code/index.html
+++ b/tour/data-model-code/index.html
@@ -153,7 +153,7 @@
 
 
   Accumulo Tour: Data Model Code
-  Tour page 4 of 9
+  Tour page 4 of 11
 
 
   Below is the solution for the exercise.
@@ -234,7 +234,7 @@ if (e.keyCode == '39') { window.location = '/tour/authorizations/'; }
     <
     
 
-    4 / 9
+    4 / 11
     
     >
     
diff --git a/tour/data-model/index.html b/tour/data-model/index.html
index 9c27982..4bcdf41 100644
--- a/tour/data-model/index.html
+++ b/tour/data-model/index.html
@@ -153,7 +153,7 @@
 
 
   Accumulo Tour: Data Model
-  Tour page 3 of 9
+  Tour page 3 of 11
 
 
   Data is stored in Accumulo in a distributed sorted map. The Keys of the map are broken up logically into a few different parts, 
@@ -202,7 +202,7 @@ if (e.keyCode == '39') { window.location = '/tour/data-model-code/'; }
     <
     
 
-    3 / 9
+    3 / 11
     
     >
     
diff --git a/tour/getting-started/index.html b/tour/getting-started/index.html
index d61c5bd..732d789 100644
--- a/tour/getting-started/index.html
+++ b/tour/getting-started/index.html
@@ -153,7 +153,7 @@
 
 

   Accumulo Tour: Getting Started
-  Tour page 1 of 9
+  Tour page 1 of 11
 
 
   First make sure you have Java, Maven and Git installed on your machine.  Oh you are already rocking? OK let’s go!
@@ -202,7 +202,7 @@ if (e.keyCode == '39') { window.location = '/tour/basic-read-write/'; }
   
     
 
-    1 / 9
+    1 / 11
     
     >
     
diff --git a/tour/index.html b/tour/index.html
index a1dda02..81c3190 100644
--- a/tour/index.html
+++ b/tour/index.html
@@ -191,6 +191,12 @@ or create an issue
     Batch Scanner Code
   
+  
+    Conditional Writer
+  
+  
+    Conditional Writer Code
+  
 
 
 
diff --git a/tour/ranges-splits/index.html b/tour/ranges-splits/index.html
index a0f74d4..e70b03c 100644
--- a/tour/ranges-splits/index.html
+++ b/tour/ranges-splits/index.html
@@ -153,7 +153,7 @@
 
 
   Accumulo Tour: Ranges and Splits
-  Tour page 7 of 9
+  Tour page 7 of 11
 
 
   A Range is a specified group of Keys. There are many different ways to create a Range.  Here are a few examples:
@@ -212,7 +212,7 @@ if (e.keyCode == '39') { window.location = '/tour/batch-scanner/'; }
     <
     
 
-    7 / 9
+    7 / 11
     
     >
     
diff --git a/tour/using-iterators/index.html b/tour/using-iterators/index.html
index f40cf94..4a970f1 100644
--- a/tour/using-iterators/index.html
+++ b/tour/using-iterators/index.html
@@ -153,7 +153,7 @@
 
 
   Accumulo Tour: Using Iterators
-  Tour page  of 9
+  Tour page  of 11
 
 
   
@@ -172,7 +172,7 @@ document.body.onkeyup = function(e){
   
     
 
-     / 9
+     / 11
     
   
 

-- 
To stop receiving notification emails like this one, please contact
['"commits@accumulo.apache.org" '].
Accumulo Tour: Authorizations Code

Accumulo Tour: Authorizations

Accumulo Tour: Writing and Reading

Accumulo Tour: Batch Scanner Code

Accumulo Tour: Batch Scanner

Ranges and Splits

Conditional Writer Code

Accumulo Tour: Ranges and Splits

Accumulo Tour: Conditional Writer Code

- < + < - 7 / 9 - - > + 11 / 11

Accumulo Tour: Conditional Writer

Exercise

+ < + - / 9 + 10 / 11 + + >

Accumulo Tour: Data Model Code

Accumulo Tour: Data Model

Accumulo Tour: Getting Started

Accumulo Tour: Ranges and Splits

Accumulo Tour: Using Iterators

- / 9 + / 11
