Question

我正在尝试找到获取所有数据砖表行数的最佳方法。这是我想出的：

for row in dvdbs.rdd.collect():
   tmp = "show tables from " + row['databaseName'] + " like 'xxx*'"
   if row['databaseName'] == 'default':
       dftbls = sqlContext.sql(tmp)
   else:
     dftbls = dftbls.union(sqlContext.sql(tmp))
tmplist = []
for row in dftbls.rdd.collect():
    tmp = 'select * from ' + row['database'] + '.' + row['tableName']
    tmpdf = sqlContext.sql(tmp)
    tmplist.append((row['database'], row['tableName'],tmpdf.count()))
columns =  ['database', 'tableName', 'rowCount']     
df = spark.createDataFrame(tmplist, columns)    
display(df)

Answer 1

我发现这要快得多...

dftbl = sqlContext.sql("show tables")
dfdbs = sqlContext.sql("show databases")
for row in dfdbs.rdd.collect():
    tmp = "show tables from " + row['databaseName'] 
    if row['databaseName'] == 'default':
        dftbls = sqlContext.sql(tmp)
    else:
       dftbls = dftbls.union(sqlContext.sql(tmp))
tmplist = []
for row in dftbls.rdd.collect():
    try:
      tmp = 'select count(*) myrowcnt from ' + row['database'] + '.' + row['tableName']
      tmpdf = sqlContext.sql(tmp)
      myrowcnt= tmpdf.collect()[0]['myrowcnt'] 
      tmplist.append((row['database'], row['tableName'],myrowcnt))
    except:
      tmplist.append((row['database'], row['tableName'],-1))

columns =  ['database', 'tableName', 'rowCount']     
df = spark.createDataFrame(tmplist, columns)    
display(df)

Answer 2

您也可以尝试使用此功能：-

def fn_byDBgetCount():
  final_list = []
  dbList = spark.sql("show databases").select("namespace").rdd.flatMap(lambda x: x).collect()
  for databaseName in dbList:
    spark.sql("use {}".format(databaseName))
    tableList = spark.sql("show tables from {}".format(databaseName)).select("tableName").rdd.flatMap(lambda x: x).collect()
    for tableName in tableList:
      tableCount = spark.sql("select count(*) as tableCount from {}".format(tableName)).collect()[0][0]
      final_list.append(list([databaseName,tableName,tableCount]))
  column_names = list(['DatabaseName','TableName','TableCount'])
  df = spark.createDataFrame(final_list,column_names)
  display(df)

fn_byDBgetCount()

azure databricks计算所有表中的行-有没有更好的方法

2 个答案: