2.3.4 Get expect scanning results
In order to have a better understanding of the below explanation, I need to explain how does a KeyValue be stored in HFile.
n How does a KeyValue be stored
Figure 4 KeyValue Storage format
We should pay special attention to Key length which present in the first part of the above picture: It’s used to store an integer value, which represents the Key length. Actually, the value is calculated by the below formula:
Ø Key length = Key infrastructure size + Actual RowKey length + Actual Column family length + Actual Qualifier length.
Ø Key infrastructure size
Size of space where stores RowKey length +
Size of space where stores Column family length +
Size of space where stores TimeStamp +
Size of space where stores Datatype
= 2 + 1 + 8 + 1 = 12.
We can see, this structure was designed without storing the qualifier length. So basing on the stored value, how to calculate the qualifier length?
Figure 5 Calculate qualifier length
n Introduction to all types of Scanners
Figure 6 Scanning flow
Ø InternalScanner can get a list of KeyValues. KeyValueScanner can only get one KeyValue. The implementation of InternalScanner is relegated to KeyValueScanner.
Ø RegionScanner,StoreScanner implement the interface of InternalScanner.
Ø StoreScanner, MemStoreScanner, StoreFileScanner implement the interface of KeyValueScanner.
Definition: If one scanner A is made up by Scanner A-1, Scanner A-2, Scaner A-3, we call the scanner A as parent-Scaner, and call Scanner A-1, Scanner A-2, Scaner A-3 as child-Scaner. This definition will be used in scope of this document.
² How to organize a group of same/similar type of scanners
A RegionScanner is made up by a group of StoreScanner. And a StoreScanner is made up by a MemStoreScanner and a group of StoreFileScanner. All those scanners are merged in a class named KeyValueHeap:
Figure 7 KeyValueHeap
There’s one member variable defined as “PriorityQueue<KeyValueScanner> heap” which used to store all the child-Scanners.
² How to switch from one scanner to another scanner
Figure 8 Switching between StoreScanners
The critical data structure is the PriorityQueue and the Comparator(The exact name is KVScannerComparator):
Ø PriorityQueue: Poll one Scanner from this queue each time, and put back after using it(If peek to the next KeyValue of this scanner is null, the scanning is finished to this scanner. So we can close it. No need to put it back.). Which Scanner should be polled determined by the below Comparator.
Ø Comparator: This is how does the comparator works:
Figure 9 KVScanenrComparator
Ø Regarding on the peek: Just look at the next KeyValue in this Scanner, but do not iterate scanner.
Switch from one StoreFileScanner to another StoreFileScanner is much easier than the above flow. Only finish one StoreFileScanner then close and switch to the next StoreFileScanenr.
² How to seek a given KeyValue from a StoreFileScanner
Seeking is widely used in Scanning and Filter. Likes the below scenarios:
Ø Start a scaning with a specified startKey. We need to seek to the nearest KeyValue larger or equal with the given startKey.
Ø One row or one columnFamily is totally filtered out. So seek to the next valid KeyValue.
This is the flow of seeking:
Figure 10 How to seek a KeyValue
² Summarize all scanners
Let’s summarize all the scanners, and see the relationships between them:
Figure 11 Relationship between all scanners
Ø KeyValueHeap: This class is used at the Region level to merge across Stores and at the Store level to merge across the memstore and StoreFiles. It is a member variable of RegionScannerImpl and StoreScanner.
· RegionScannerImpl use KeyValueHeap to merge across Stores.
· StoreScanenr use KeyValueHeap to merge across MemStore and StoreFiles.
So, we can say KeyValueHeap is a parent-Scanner. It has one or more child-Scanners. During the instantiation of this class, it will load all the child-Scanners.
Ø KeyValueScanner: It is used to get the next KeyValue. Meanwhile, it can be used to look up the next KeyValue(Without iterate scanner), seek a KeyValue.
Ø InternalScanner: It is used to get number of KeyValues(We can specify how many KeyValues we get by setting the parameter of batch. Default is get all the KeyValues from one row.)
1.1.1 Close RegionScanner
The closure flow is simple:
Figure 12 Closure flow
- •python书籍推荐：Head First Python（中文版）
- •gentoo ‘/usr/bin/python’ is not valid symlink
- •24.1. Tkinter — Python interface to Tcl/Tk — Python v2.7.2 documentation 24.1. Tkinter — Python int…
- •MySQLdb for Python使用指南/Python的数据库操作
- •Python 3.0 最全学习02——input / 字符串拼接 / ifelse
- •Linux – yum错误：-bash: /usr/bin/yum: /usr/bin/python: bad interpreter: No such file or directory
- •Python -c / -m 执行命令
- •Kafka Java SSL(Consumer/Producer) Demo