删除所有 Azure 表记录

Question

我有一个 Azure 存储表，它有 3k+ 条记录。

删除表中所有行的最有效方法是什么？

Answer 1

对于 3000 条记录，最简单的方法是 delete the table。但是请注意，当您删除表时，它并没有在当时被删除，而是被放入某种待删除的队列中，并且实际上是在一段时间后被删除的。这个时间取决于系统的负载 + 表中的实体数量。在此期间，您将无法重新创建此表或使用此表。

如果继续使用该表对您很重要，则唯一的其他选择是删除实体。为了更快地删除，您可以查看使用Entity Batch Transactions 删除实体。但要删除实体，您需要先获取实体。您可以通过仅获取实体的PartitionKey 和RowKey 属性而不是获取所有属性来加快获取过程，因为删除实体只需要这两个属性。

Answer 2

我使用这样的东西。我们按日期对键进行分区，您的情况可能会有所不同：

async Task Main()
{
    var startDate = new DateTime(2011, 1, 1);
    var endDate = new DateTime(2012, 1, 1);

    var account = CloudStorageAccount.Parse("connString");
    var client = account.CreateCloudTableClient();
    var table = client.GetTableReference("TableName");

    var dates = Enumerable.Range(0, Math.Abs((startDate.Month - endDate.Month) + 12 * (startDate.Year - endDate.Year)))
        .Select(offset => startDate.AddMonths(offset))
        .ToList();

    foreach (var date in dates)
    {
        var key = $"{date.ToShortDateString()}";

        var query = $"(PartitionKey eq '{key}')";
        var rangeQuery = new TableQuery<TableEntity>().Where(query);

        var result = table.ExecuteQuery<TableEntity>(rangeQuery);
        $"Deleting data from {date.ToShortDateString()}, key {key}, has {result.Count()} records.".Dump();

        var allTasks = result.Select(async r =>
        {
            try
            {
                await table.ExecuteAsync(TableOperation.Delete(r));
            }
            catch (Exception e) { $"{r.RowKey} - {e.ToString()}".Dump(); }
        });
        await Task.WhenAll(allTasks);
    }
}

Answer 3

这取决于您的数据结构，但如果您可以为所有记录编写查询，则可以将每个记录添加到 TableBatchOperation 并一次全部执行。

这是一个仅在同一分区键中获取所有结果的示例，改编自 How to get started with Azure Table storage and Visual Studio connected services。

// query all rows
CloudTable peopleTable = tableClient.GetTableReference("myTableName");
var query = new TableQuery<MyTableEntity>();
var result = await remindersTable.ExecuteQuerySegmentedAsync(query, null);

// Create the batch operation.
TableBatchOperation batchDeleteOperation = new TableBatchOperation();

foreach (var row in result)
{
    batchDeleteOperation.Delete(row);
}

// Execute the batch operation.
await remindersTable.ExecuteBatchAsync(batchDeleteOperation);

Answer 4

我使用下面的函数，先将所有分区键放入队列，然后循环遍历键，以100个为单位批量删除所有行。

Queue queue = new Queue();
            queue.Enqueue("PartitionKeyTodelete1");
            queue.Enqueue("PartitionKeyTodelete2");
            queue.Enqueue("PartitionKeyTodelete3");

            while (queue.Count > 0)
            {
                string partitionToDelete = (string)queue.Dequeue();

                TableQuery<TableEntity> deleteQuery = new TableQuery<TableEntity>()
                  .Where(TableQuery.GenerateFilterCondition("PartitionKey", QueryComparisons.Equal, partitionToDelete))
                  .Select(new string[] { "PartitionKey", "RowKey" });

                TableContinuationToken continuationToken = null;

                do
                {
                    var tableQueryResult = await myTable.ExecuteQuerySegmentedAsync(deleteQuery, continuationToken);

                    continuationToken = tableQueryResult.ContinuationToken;

                    // Split into chunks of 100 for batching
                    List<List<TableEntity>> rowsChunked = tableQueryResult.Select((x, index) => new { Index = index, Value = x })
                        .Where(x => x.Value != null)
                        .GroupBy(x => x.Index / 100)
                        .Select(x => x.Select(v => v.Value).ToList())
                        .ToList();

                    // Delete each chunk of 100 in a batch
                    foreach (List<TableEntity> rows in rowsChunked)
                    {
                        TableBatchOperation tableBatchOperation = new TableBatchOperation();
                        rows.ForEach(x => tableBatchOperation.Add(TableOperation.Delete(x)));

                        await myTable.ExecuteBatchAsync(tableBatchOperation);
                    }
                }
                while (continuationToken != null);
            }

Answer 5

对于后来发现这一点的人来说，接受答案“刚刚删除表”的问题在于，虽然它在存储模拟器中运行良好，但在生产中会随机失败。如果您的应用/服务需要定期重新生成表，那么您会发现由于冲突或删除仍在进行中而导致失败。

相反，我发现最快和最防错的 EF 友好方法是删除分段查询中的所有行。下面是我正在使用的一个简单的嵌入式示例。传入您的客户端、表名和实现 ITableEntity 的类型。

private async Task DeleteAllRows<T>(string table, CloudTableClient client) where T: ITableEntity, new()
    {
        // query all rows
        CloudTable tableref = client.GetTableReference(table);           
        var query = new TableQuery<T>();
        TableContinuationToken token = null;
                                         
        do
        {
            var result = await tableref.ExecuteQuerySegmentedAsync(query, token);  
            foreach (var row in result)
            {
                var op = TableOperation.Delete(row);
                tableref.ExecuteAsync(op);
            }
            token = result.ContinuationToken;
        } while (token != null);  
        
    }

示例用法：

table = client.GetTableReference("TodayPerformanceSnapshot");
created = await table.CreateIfNotExistsAsync();

if(!created)
{ 
    // not created, table already existed, delete all content
   await DeleteAllRows<TodayPerformanceContainer>("TodayPerformanceSnapshot", client);
   log.Information("Azure Table:{Table} Purged", table);
}

批处理方法需要付出更多的努力，因为您必须同时处理“批处理中只有相同的分区键”和“只有 100 行”的限制。以下版本的 DeleteAllRows 执行此操作。

private async Task DeleteAllRows<T>(string table, CloudTableClient client) where T: ITableEntity, new()
    {
        // query all rows
        CloudTable tableref = client.GetTableReference(table);           
        var query = new TableQuery<T>();
        TableContinuationToken token = null;            
        TableBatchOperation batchops = new TableBatchOperation();
        Dictionary<string, Stack<TableOperation>> pendingOperations = new Dictionary<string, Stack<TableOperation>>();
        
        do
        {
            var result = await tableref.ExecuteQuerySegmentedAsync(query, token);
            foreach (var row in result)
            {
               var op = TableOperation.Delete(row);
                if (pendingOperations.ContainsKey(row.PartitionKey))
                {
                    pendingOperations[row.PartitionKey].Push(op);
                }
                else
                {
                    pendingOperations.Add(row.PartitionKey, new Stack<TableOperation>() );
                    pendingOperations[row.PartitionKey].Push(op);
                }                                    
            }
            token = result.ContinuationToken;
        } while (token != null);

        // order by partition key            
        foreach (var key in pendingOperations.Keys)
        {                
            log.Information($"Deleting:{key}");                
            var rowStack = pendingOperations[key];
            int max = 100;
            int current = 0;

            while (rowStack.Count != 0)
            {
                // dequeue in groups of 100
                while (current < max && rowStack.Count > 0)
                {
                    var op = rowStack.Pop();
                    batchops.Add(op);
                    current++;
                }

                //execute and reset
                _ = await tableref.ExecuteBatchAsync(batchops);
                log.Information($"Deleted batch of size:{batchops.Count}");
                current = 0;
                batchops.Clear();
            }
        }                       
    }

Answer 6

我最近编写了一个可以做到这一点的库。

来源/文档：https://github.com/pflajszer/AzureTablesLifecycleManager

对于您的用例，代码如下所示：

// inject ITableManager in the constructor:

private readonly ITableManager _api;

public MyClass(ITableManager api)
{
    _api = api;
}

/// <summary>
/// Delete all data from a single table
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="tableName"></param>
/// <returns></returns>
public Task<DataTransferResponse<T>> DeleteTableDataAsync<T>(string tableName) where T : class, ITableEntity, new()
{
    // this query will return a single table with a given name:
    Expression<Func<TableItem, bool>> tableQuery = x => x.Name == tableName;

    // this query will return all the data from the table:
    Expression<Func<T, bool>> dataQuery = x => true;
             
    // ... but you can use LINQ to filter results too, like:
    // Expression<Func<T, bool>> anotherExampleOfdataQuery = x => x.Timestamp < DateTime.Now.AddYears(-1);

    return _api.DeleteDataFromTablesAsync<T>(tableQuery, dataQuery);
}

...或者，正如 Gaurav Mantri 建议的那样，您可以删除表本身：

/// <summary>
/// Delete a single table
/// </summary>
/// <param name="tableName"></param>
/// <returns></returns>
public Task<DataTransferResponse<TableItem>> DeleteTableAsync(string tableName)
{
    // this query will return a single table with a given name:
    Expression<Func<TableItem, bool>> tableQuery = x => x.Name == tableName;

    return _api.DeleteTablesAsync(tableQuery);
}