Update hands-on.ipynb

3c7b500f · Chao Zhang · 5e5730e6 · 3c7b500f
Commit 3c7b500f authored 4 years ago by Chao Zhang
--- a/hands-on.ipynb
+++ b/hands-on.ipynb
@@ -359,7 +359,7 @@
    "### 4.2 Hands-on exercises\n",
    "  There are five tasks for querying the multi-model data as follows:\n",
    "  \n",
-    "  **Q1: Get the top-10 best-selling products in all orders.** Hint: We assume the quantity of each product in the orderline is one. Use the wildcard [*] to access the Orderline array, then use the flatten operator to expand the sub-array of productId, finally collect the ids into groups.\n",
+    "  **Q1: Get the top-10 best-selling products in all orders.** Hint: we assume the quantity of each product in the orderline is one. Use the wildcard [*] to access the Orderline array, then use the flatten operator to expand the sub-array of productId, finally collect the ids into groups.\n",
    "  \n",
    "  **Q2: Calculate the total cost of female’s orders in year 2008.**  Hint: involve a join between the Person table and Order files. OrderDate contains the year information. \n",
    "  \n",

 %% Cell type:markdown id: tags:

 # Hands-On Session for the tutorial: Multi-model Data query languages and processing paradigms in CIKM 2020

 %% Cell type:markdown id: tags:

 ## 1: ArangoDB Installation

 %% Cell type:markdown id: tags:

 To better follow this hands-on session, please install the ArangoDB in advance.

 You can install the lateset version by following the official instructions if your computer satisfies the requirement of v3.7.0:

 * https://www.arangodb.com/docs/stable/installation.html

 Or you may download and install the previous community builds (e.g., v3.4.0 https://download.arangodb.com/arangodb34/index.html) of ArangoDB

 We will use the ArangoDB WebUI to perform the queries, the default url is *localhost:8529*, and the default username is root with the empty password.

 %% Cell type:markdown id: tags:

 ## 2 Document store in ArangoDB : collections and documents

 *Relational databases* contain *tables* of *records* (as *rows*).

 An **ArangoDB document database** contains **collections** that contain **documents**. The documents follow the JSON format, and are usually stored in a binary format.

 Below is an example of json document containing information of a student and corresponding scores.

 ** Score Document**
 ```
 {"_id":0,"name":"aimee Zank",
 "scores":[{"score":1.463179736705023,"type":"exam"},
           {"score":11.78273309957772,"type":"quiz"},
           {"score":35.8740349954354,"type":"homework"}]
 }
 ```

 %% Cell type:markdown id: tags:

 ### 2.1  Warm-up: Loading the score documents

 ### The score file is kept here. https://version.helsinki.fi/chzhang/cikm-2020-hands-on-session-for-multi-model-queries/-/blob/master/scores.json

 Please import the file to the server using arangoimport as follows:

 ./arangoimport --file PATH-TO/scores.json --collection scores --create-collection true

 Notes:

 arangoimport options: https://www.arangodb.com/docs/stable/programs-arangoimport-options.html

 I recommend to add the arangoimport path to the system PATH variable:

 Default Path in Mac for arangoimport: /Applications/ArangoDB3-CLI.app/Contents/Resources

 Default Path in Linux for arangoimport: /usr/local/Cellar/arangodb

 Default Path in Windows for arangoimport: C:\Program Files\ArangoDB

 %% Cell type:markdown id: tags:

 ### 2.2 Querying the documents

 %% Cell type:code id: tags:

 ``` python
 # Query 1: return a score document in the collection.

 For doc in scores Filter doc.name =="Leonida Lafond" return doc
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 2: (multiple conditions) return a score document in the collection.
 For doc in scores Filter doc.name =="Leonida Lafond" and doc._key=='@key' return doc
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 3: (array operator 1) find types of scores.
 For doc in scores limit 1 return doc.scores[*].type
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 4: (array operator 2) find students whose exam scores are greater than 90.
 For doc in scores limit 1 return doc.scores[* Filter CURRENT.score>90].score
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 5: (array operator 3) compute the average score.
 For doc in scores limit 1 return AVERAGE(doc.scores[*].score)
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 6: flatten
 Return FLATTEN([ 1, 2, [ 3, 4 ], 5, [ 6, 7 ], [ 8, [ 9, 10 ] ] ])
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 7: sorting
 For doc in scores
    Sort first(doc.scores[*].score) DESC
    Return doc
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 8: grouping (with or without count)
 For doc in scores
    COLLECT name=doc.name into g
    return {name,g}
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 9: define a variable using Let
 FOR doc in scores
    LET average_score=AVERAGE(doc.scores[*].score)
    SORT average_score DESC
    RETURN { name:doc.name,average_score:average_score}
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 10: Inner join between two collections
    FOR doc1 in scores
        FOR doc2 in scores
            Filter doc1.name==doc2.name
            return {doc1:doc1,doc2:doc2}
 ```

 %% Cell type:markdown id: tags:

 ## 3. Graph store : nodes and edges

 An ArangoDB graph database contains a set of node collections and edge collections.

 %% Cell type:markdown id: tags:

 ### 3.1 Loading the example graphs

 %% Cell type:markdown id: tags:

 arangosh> var examples = require("@arangodb/graph-examples/example-graph.js");
 arangosh> var g = examples.loadGraph("knows_graph");

 %% Cell type:markdown id: tags:

 ### 3.2 Traversing the graphs

 FOR vertex[, edge[, path]]
  IN [min[..max]]
  OUTBOUND|INBOUND|ANY startVertex
  GRAPH graphName
  [OPTIONS options]

 %% Cell type:code id: tags:

 ``` python
 # Query 11: find the friends of a given person.

 // get a random person p
 Let p= (For person in persons Sort rand() limit 1 return person)

 // find the friends of p
 FOR v,e,path
 IN  1..1 any p[0]._id
 GRAPH "knows_graph"
 RETURN {p,v,e}
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 12: Filtering
 # Filtering vertex
 // get person bob
 Let p= (For person in persons Filter person._key=='bob' return person)

 // find the friends of p
 FOR v,e
 IN  1..1 any p[0]._id
 GRAPH "knows_graph"
 Filter v._key=='alice'
 RETURN {p,v,e}

 # Filtering path
 // get person bob
 Let p= (For person in persons Filter person._key=='bob' return person)

 // find the friends of p
 FOR v,e,path
 IN  1..2 any p[0]._id
 GRAPH "knows_graph"
 Filter length(path.edges)>1
 RETURN {p,v,e,path}
 ```

 %% Cell type:code id: tags:

 ``` python
 # Query 13: Graph functions -- Shortest Path

 // find the friends of p
 FOR v,e
 IN Any SHORTEST_PATH
 'persons/charlie' to 'persons/alice'
 GRAPH "knows_graph"
 RETURN {v,e}
 ```

 %% Cell type:markdown id: tags:

 ## 4. Your turn - Querying the multi-model datasets

 %% Cell type:markdown id: tags:

 ### 4.1 Data description
 The data consists of three files: KnowsGraph.csv, Order.json, and Person.csv.  Specifically, Person is the tabular data with fields(id, firstName, lastname, gender, birthday, creationDate, locationIp, browserUsed). KnowsGraph is the linked data, each row is an edge starting from a PersonId to another PersonId. Order is the json data with fields (OrderId, PersonId, OrderDate,TotalPrice, [Orderline:[productId,title,price ]). Note that Orderline is an array including more than one product.

 Download the data here https://version.helsinki.fi/chzhang/cikm-2020-hands-on-session-for-multi-model-queries/-/tree/master/Multi-model-data and import them using arangoimport as follows:

 ./arangoimport --file PATH-TO/Multi-model-data/Person.csv --type csv --translate "id=_key" --collection "Person" --server.username root --create-collection true

 ./arangoimport --file PATH-TO/Multi-model-data/Order.json --type json --translate "id=_key" --collection "Order" --server.username root --create-collection true

 ./arangoimport --file PATH-TO/Multi-model-data/KnowsGraph.csv --type csv --translate "from=_from" --translate "to=_to" --collection "KnowsGraph" --from-collection-prefix Person --to-collection-prefix Person --server.username root --create-collection true --create-collection-type edge

 %% Cell type:markdown id: tags:

 ### 4.2 Hands-on exercises
  There are five tasks for querying the multi-model data as follows:

-  **Q1: Get the top-10 best-selling products in all orders.** Hint: We assume the quantity of each product in the orderline is one. Use the wildcard [*] to access the Orderline array, then use the flatten operator to expand the sub-array of productId, finally collect the ids into groups.
+  **Q1: Get the top-10 best-selling products in all orders.** Hint: we assume the quantity of each product in the orderline is one. Use the wildcard [*] to access the Orderline array, then use the flatten operator to expand the sub-array of productId, finally collect the ids into groups.

  **Q2: Calculate the total cost of female’s orders in year 2008.**  Hint: involve a join between the Person table and Order files. OrderDate contains the year information.

  **Q3: Given a start person (_key='2199023262543'), return the number of orders made by this person’s friends in 2009.** Hint: friends are the outbound vertices of the start person.

  **Q4: Given PersonX (_key='2199023259756') and PersonY (_key='26388279077535'), find the shortest path between them, and also return TOP-5 best-selling products for all persons in that path including PersonX and PersonY.** Hint: use SHORTEST_PATH function, see Query 13 as an example.

  **Q5: Find the top-2 persons who spend the highest amount of money in JSON orders. Then for each person, traverse her knows-graph with 3-hop to find the friends, and finally return the number of common friends for these two persons.** Hint: use INTERSECTION function to find the common items of two lists.



 %% Cell type:code id: tags:

 ``` python
 ```