!pip install pymongo kafka-python --quiet

接下来，我们将连接到 MongoDB Atlas 和 Kafka 代理：

from kafka import KafkaConsumer
from pymongo import MongoClienttry:client = MongoClient("mongodb+srv://<username>:<password>@<host>/?retryWrites=true&w=majority")db = client.adtechprint("Connected successfully")
except:print("Could not connect")consumer = KafkaConsumer("ad_events",bootstrap_servers = ["public-kafka.memcompute.com:9092"]

我们将用我们之前从 MongoDB Atlas 保存的值替换<username>,<password>和。<host>

SHOW TABLES;

这应该显示一张名为events：

+------------------+
| Tables_in_adtech |
+------------------+
| events           |
+------------------+

我们将检查表的结构：

DESCRIBE events;

输出应如下所示：

+-------+------+------+------+---------+-------+
| Field | Type | Null | Key  | Default | Extra |
+-------+------+------+------+---------+-------+
| _id   | text | NO   | UNI  | NULL    |       |
| _more | JSON | NO   |      | NULL    |       |
+-------+------+------+------+---------+-------+

接下来，我们将检查是否有pipelines：

SHOW PIPELINES;

这将显示events当前调用的一个管道Stopped：

+---------------------+---------+-----------+
| Pipelines_in_adtech | State   | Scheduled |
+---------------------+---------+-----------+
| events              | Stopped | False     |
+---------------------+---------+-----------+

现在我们将启动events管道：

START ALL PIPELINES;

并且状态应更改为Running：

+---------------------+---------+-----------+
| Pipelines_in_adtech | State   | Scheduled |
+---------------------+---------+-----------+
| events              | Running | False     |
+---------------------+---------+-----------+

如果我们现在运行以下命令：

SELECT COUNT(*) FROM events;

它应该返回 100 作为结果：

+----------+
| COUNT(*) |
+----------+
|      100 |
+----------+

我们将检查表中的一行events，如下所示：

SELECT * FROM events LIMIT 1;

输出应类似于以下内容：

+--------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| _id                                  | _more                                                                                                                                                                                                                                                                   |
+--------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| {"$oid": "64ec906d0e8c0f7bcf72a8f7"} | {"_id":{"$oid":"64ec906d0e8c0f7bcf72a8f7"},"advertiser":"Wendys","campaign":13,"country":"US","event_name":"Click","gender":"Female","income":"75k - 99k","page_url":"/2014/05/flamingo-pop-bridal-shower-collab-with.html","region":"New Mexico","user_id":3857963416} |
+--------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

CDC 解决方案已成功连接到 MongoDB Atlas 并将所有 100 条记录复制到 SingleStore Kai。

SELECT COUNT(*) FROM events;

SELECT _more::country AS `events.country`, COUNT(_more::country) AS 'events.countofevents'
FROM adtech.events AS events
GROUP BY 1;

SELECT _more::advertiser AS `events.advertiser`, COUNT(*) AS `events.count`
FROM adtech.events AS events
WHERE (_more::advertiser LIKE '%Subway%' OR _more::advertiser LIKE '%McDonals%' OR _more::advertiser LIKE '%Starbucks%' OR _more::advertiser LIKE '%Dollar General%' OR _more::advertiser LIKE '%YUM! Brands%' OR _more::advertiser LIKE '%Dunkin Brands Group%')
GROUP BY 1
ORDER BY `events.count` DESC;

SELECT *
FROM (SELECT *, DENSE_RANK() OVER (ORDER BY xx.z___min_rank) AS z___pivot_row_rank, RANK() OVER (PARTITION BY xx.z__pivot_col_rank ORDER BY xx.z___min_rank) AS z__pivot_col_ordering, CASEWHEN xx.z___min_rank = xx.z___rank THEN 1ELSE 0END AS z__is_highest_ranked_cellFROM (SELECT *, Min(aa.z___rank) OVER (PARTITION BY aa.`events.income`) AS z___min_rankFROM (SELECT *, RANK() OVER (ORDER BY CASEWHEN bb.z__pivot_col_rank = 1 THEN (CASEWHEN bb.`events.count` IS NOT NULL THEN 0ELSE 1END)ELSE 2END, CASEWHEN bb.z__pivot_col_rank = 1 THEN bb.`events.count`ELSE NULLEND DESC, bb.`events.count` DESC, bb.z__pivot_col_rank, bb.`events.income`) AS z___rankFROM (SELECT *, DENSE_RANK() OVER (ORDER BY CASEWHEN ww.`events.gender` IS NULL THEN 1ELSE 0END, ww.`events.gender`) AS z__pivot_col_rankFROM (SELECT _more::gender AS `events.gender`, _more::income AS `events.income`, COUNT(*) AS `events.count`FROM adtech.events AS eventsWHERE (_more::income <> 'unknown' OR _more::income IS NULL)GROUP BY 1, 2) ww) bbWHERE bb.z__pivot_col_rank <= 16384) aa) xx) zz
WHERE (zz.z__pivot_col_rank <= 50 OR zz.z__is_highest_ranked_cell = 1) AND (zz.z___pivot_row_rank <= 500 OR zz.z__pivot_col_ordering = 1)
ORDER BY zz.z___pivot_row_rank;

图 2 显示了 AdTech 仪表板上图表大小和位置的示例。我们将自动刷新选项设置为 1 分钟。