At Boxalino we are strong advisors for ELT flows (vs ETL). The major benefit is speed, transparency and maintenence: data is loaded directly into a destination system (BQ), and transformed in-parallel (Dataform).
The ELT flow is recommended when wanting to load your data as is (no transformation) in BQ and then perform the transformation logic (SQL) in Dataform. Integration-effort is minimal (push all data to GCS - Boxalino-owned or private client-owned, load data to BQ, write SQL in Dataform).
More information about the Integration Strategies is available on Confluence.
Integration Strategies
Data Integration (DI) Generic Flow
The Generic Data Integration (DI) Flow expects for the client, themselves to execute the steps described in Data Integration .
For certain platforms (Shopware6, Magento2), the step #1 can be done with a Boxalino Data Integration plugin.
For generic clients, the JSONL generation is managed by the clients` team by following the Boxalino Data Structure requirements Data Structure (as documented).
The Boxalino Data Structure and integration flows/traits are publicly available as a PHP repository https://github.com/boxalino/data-integration-doc-php
As a integrator, you can do the steps #1-#4 Data Integration by following our recommended integration flowIntegration Flow
Data Integration (DI) SAAS Flow
The di-saas service is designed to:
be used for custom integrations by ingesting data (.csv, .jsonl) from SFTP, GCS, BQ or public links.
integrate data from targetted platforms (ex: PlentyMarkets)
For both cases, a mapping is done between the Boxalino Data Structure Data Structure and the input sources (either files or API endpoints) with the help of BQ & Dataform.
In the case of a custom integration, the di-saas flow is composed of the following scopes:
Data connector & Data load
downloads the files from remote source to GCS
loads the raw content to the transform dataset (<account>_T)
Data transformation (with logic/SQL provided by the client)
Steps #1-#4 from the Generic Flow
DI automation
DI-SAAS Request (Overview)
The DI (data integration) request has all the required information in a JSON body.
The elements are for connector (+ files load options) and the di (data integration) request parameters (default configurations).
[
{
"connector": {
"type": "sftp|gcs|plentymarket|plytix|boxalino",
"options": {
// specific for each connector type
},
"load": {
"options": {
// for loading the data in BQ (BQ parameters)
"format": "CSV|NEWLINE_DELIMITED_JSON",
"field_delimiter": ";",
"autodetect": true,
"schema": "",
"skipRows": 1(CSV)|0(JSONL),
"max_bad_records": 0,
"quote": "",
"write_disposition": "WRITE_TRUNCATE",
"create_disposition": "",
"encoding": ""
},
"doc_X": {
"property_node_from_doc_data_structure": [
{
"source": "file___NODASHDATE__.jsonl",
"name": "<used to create suffix for BQ table>",
"schema": ""
}
]
}
}
},
"di": {
"configuration": {
"languages": [
"de",
"fr"
],
"currencies": [
"CHF"
],
"mapping": {
"languagesCountryCode": {"de":"CH_de", "fr":"CH_fr"}
},
"default": {
"language": "de",
"currency": "CHF"
}
}
}
}
]
1. Data Connector & Data Load
The access to the data is managed by the client. These are a few guidelines for naming patterns:
The relevant data sources are available in .csv or JSONL (prefered) format
The files have a timestamp in the naming or in the access path (ex: product_20221206.jsonl)
this will help automating the integration
The files required to update a certain data type (ex: product, order, etc) are available in the same path
The files are available on an endpoint (SFTP, GCS, 3rd party API, public URL) to which Boxalino has access
for GCS/ GCP sources: access being shared to Boxalino`s Service Account
boxalino-di-api@rtux-data-integration.iam.gserviceaccount.comfor AWS / SFTP : the client`s own AWS/SFTP environment with a Boxalino user & credentials
The load configuration defines the GCP properties for loading content to BQ.
The requirements specified above (#1-#4) are necessary if the data is accessed from a remote (outside Boxalino) scope.
If your integration exports the data directly in Boxalino (as described https://boxalino-internal.atlassian.net/wiki/spaces/DOC/pages/2606792705/Boxalino+Data+Integration+DI-SAAS+-+ELT+Flow#Loading-content-to-Boxalino-GCS-(connector%3A-boxalino) ), please continue with the Data Transformation step.
It is possible to configure dynamic source names (to identify the files loaded in GCS) (ex: if the content to be loaded in BQ is exported in batches or has a dynamic suffix).
The supported variations are:__DATE__ (Y-m-d), __NODASHDATE__ (Ymd), __TM__ (Y-m-dTH:i:s), __NODASHTM__ (YmdHis), __DATE_PATH__ (Y/m/d), __JN_DATE_PATH__ (Y/n/j), __DATE_YESTERDAY__ (Y-m-d -1 day), __DATE_MONTH__ (Y-m-01), __NODASHDATE_MONTH__ (Ym01), __NODASHDATE_YESTERDAY__ (Ymd -1 day).
For example, if there are 3 chunks for order-<tm>-1.jsonl, the configuration for load can be:"load":{"options":{<add the default BQ-load options}, "doc_order":{"entity":[{"source":"doc_order___NODASHTM__*.jsonl}]}}
2. Data Transformation
Once the data is loaded in GCS and BQ, it is time to transform it in the necessary data structure.
The transformation happens by preparing a BQ SQL for every node part of the doc_X data structure (ex: title, stock, price, string_attributes, etc). The output of the SQL will be of a certain format/structure for each property.
As this step was done in-house by Boxalino, for the POC of our ELT solution, further definition will be provided.
Dataform
The transformation happens with the help of Google Dataform https://cloud.google.com/dataform .
This implies the following:
The client has access to a GCP project
The client will create a Dataform repository https://cloud.google.com/dataform/docs/repositories
The client has access to a GitHub or GitLab repository (to connect it to the Dataform repository) https://cloud.google.com/dataform/docs/connect-repository
The client has given “Dataform Admin” permission to Boxalino Service Account
boxalino-dataform@rtux-data-integration.iam.gserviceaccount.com
When using dataform for transforming the exported data (your custom CSV/JSONL files) into Boxalino Data Structure, the JSON body (for the SYNC REQUEST) has the following information (next to connector and di):
"transform": {
"vars": {
"var1": "value"
},
"tags": [
"order"
],
"dataform": {
"project": "rtux-data-integration",
"location": "europe-west1",
"repository": "boxalino-di-saas-elt",
"workspace": "dataform"
}
}
The values for dataform are available directly in your dataform project link:https://console.cloud.google.com/bigquery/dataform/locations/<location>/repositories/<repository>/workspaces/<workspace>?project=<project>
The DI-SAAS SYNC request
The DI request will use the same headers (client, tm, mode, type, authorization) and a JSON request body that would provide mapping details between the loaded .jsonl files and data meaning.
it will load the files to BQ in a T (transform) dataset (ex:
SELECT * FROM rtux-data-integration.<account>_T.<tm>_<doc_X>_<property>_<name>)it will run the transform flow (step #1-#3 from Data Integration )
generating each doc_X JSONL content
loading the doc_X JSONL in your GCS bucket
loading the doc_X JSONL to BQ
it will run the SYNC request Sync Request for the process
loads the content in core tables in BQ
loads the content in the respective data index (for products)
REQUEST DEFINITION
As an integrator, please create the bellow request to the provided endpoint.
There should be a process within your own project that triggers the data sync between a 3rd party source (connector) and Boxalino.
Endpoint | production | ||
|---|---|---|---|
| 1 | Action | /sync | |
| 2 | Method | POST | |
| 3 | Body | ||
| 4 | Headers | Authorization | Basic base64<<DATASYNC API key : DATASYNC API Secret>> note: use the API credentials from your Boxalino account that have the ADMIN role assigned  |
| 5 |
| Content-Type | application/json |
| 6 |
| client | account name |
| 7 |
| mode | data sync mode: F for full, D for delta, E for enrichments |
| 8 |
| type | product, user, content, user_content, order. if left empty - it will check for all tables with the given tm |
| 9 |
| tm | (optional) time , in format: YmdHis; if the data was loaded in Boxalino GCS (https://boxalino.atlassian.net/wiki/spaces/BPKB/pages/928874497/DI-SAAS+ELT+Flow#Loading-content-to-Boxalino-GCS-(connector%3A-boxalino) ) - must re-use the tm from the load step technical: used to identify the documents version |
| 10 |
| ts | (optional) timestamp, must be millisecond based in UNIX timestamp |
| 11 |
| dev | (optional) use this to mark the content for the dev data index |
The DI-SAAS CORE request
The CORE request is used to facilitate the load of data (as is) in the clients` _core dataset from the Boxalino ecosystem.
In order to achieve this, there are 2 requests to be made:
load data to GCS https://boxalino.atlassian.net/wiki/spaces/BPKB/pages/928874497/DI-SAAS+ELT+Flow#Loading-content-to-Boxalino-GCS-(connector%3A-boxalino)
load data from GCS to
<client>_core.<destination>table
REQUEST DEFINITION
As an integrator, please create the bellow request to the provided endpoint.
There should be a process within your own project that triggers the data sync between a 3rd party source (connector) and Boxalino.
Endpoint | production | ||
|---|---|---|---|
| 1 | Action | /core | |
| 2 | Method | POST | |
| 3 | Body | ||
| 4 | Headers | Authorization | Basic base64<<DATASYNC API key : DATASYNC API Secret>> note: use the API credentials from your Boxalino account that have the ADMIN role assigned |
| 5 |
| Content-Type | application/json |
| 6 |
| client | account name |
| 7 |
| mode | data sync mode: F for full, D for delta, E for enrichments |
| 8 |
| type | product, user, content, user_content, order. if left empty - it will check for all tables with the given tm |
| 9 |
| tm | (optional) time , in format: YmdHis; if the data was loaded in Boxalino GCS (https://boxalino.atlassian.net/wiki/spaces/BPKB/pages/928874497/DI-SAAS+ELT+Flow#Loading-content-to-Boxalino-GCS-(connector%3A-boxalino) ) - must re-use the tm from the load step technical: used to identify the documents version |
| 10 |
| dev | (optional) use this to mark the content for the dev data index |
Loading content to Boxalino GCS (connector: boxalino)
By following these steps, you can push your data (JSONL or CSV) directly in your client`s GCS bucket in the Boxalino scope. After all data has been loaded in GCS, the DI-SAAS request can be called https://boxalino.atlassian.net/wiki/spaces/BPKB/pages/928874497/DI-SAAS+ELT+Flow#The-DI-request , assigning connector → type : boxalino.
There are 2 available flows, based on the size of your data:
The content is exported as the
bodyof yourPOSTrequestThe content is exported with the help of a public GCS Signed URL (https://cloud.google.com/storage/docs/access-control/signed-urls )
Option #1 is recommended for data volume less than 32MB.
Option #2 is allowed for any data size.
The described steps can be followed if you want to export your data in Boxalino GCP project, in your clients` account.
The LOAD request will create a GCS file in your project`s GCS buckte: gs://prod_rtux-data-integration_<account>/<type>/<tm>/<mode>_<doc>
there is no content validation at this step
When loading data to be transformed, make sure to use the T (transform) mode. This will ensure that the data is loaded in BQ in the transform datasets (<client>_T), which is available only 24h.
A. Loading content less than 32 MB
Files under 32MB can be loaded directly as REQUEST BODY CONTENT in Boxalino`s GCS.
The sample request bellow would create the languagefeed_20220317.csv in your GCS bucket gs://prod_rtux-data-integration_<account>/<type>/<tm>/<mode>_<filename>
curl "https://boxalino-di-stage-krceabfwya-ew.a.run.app/transformer/load" \ -X POST \ -H "Content-Type: application/json" \ -H "client: <account>" \ -H "dev: true|false" \ -H "tm: YYYYmmddHHiiss" \ -H "type: product|content|order|user|communication_history|communication_planning|user_generated_content" \ -H "mode: T" \ -H "doc: <filename>" \ -d "<JSONL>" \ -H "Authorization: Basic <encode of the account>"
Step #1 must be repeated for every file that is required to be added for the given process (same tm, mode & type)
The same tm value must be used across your other requests. This identifies the timestamp of your computation process.
If the service response is an error like: 413 Request Entity Too Large - please use the 2nd flow.
B. Loading undefined data size
For content over 32MB, we provide an endpoint to access a Signed GCS Url that would put all your streamed content into a file (currently there is no defined file size limit in GCS)
Read more about Google Cloud Signed URL https://cloud.google.com/storage/docs/access-control/signed-urls (response samples, uses, etc)
1. Make a request for public upload link
This is the generic POST request:
curl --connect-timeout 60 --max-time 300 "https://boxalino-di-stage-krceabfwya-ew.a.run.app/transformer/load/url" \ -X POST \ -H "Content-Type: application/json" \ -H "client: <account>" \ -H "dev: true|false" \ -H "tm: YYYYmmddHHiiss" \ -H "type: product|content|order|user|communication_history|communication_planning|user_generated_content" \ -H "mode: T" \ -H "chunk: <id>" \ -H "doc: <filename>" \ -H "Authorization: Basic <encode of the account>"
The response will be an upload link that can only be used in order to create the file in the clients` GCS bucket. The link is valid for 30 minutes.
2. Upload the content on the public link
curl --connect-timeout 60 --timeout 0 <GCS-signed-url> \
-X PUT \
-H "Content-Type: application/octet-stream" \
-d "<YOUR DOCUMENT JSONL CONTENT (STREAM)>"
The use of the header chunk is required if the same file/document is exported in batches/sections.
Repeat steps 1+2 for every data batch loaded in GCS.
Make sure to increment the value of the chunk property for each /transformer/load/url request.
Step #1 - #2 must be repeated for every file that is required to be added for the given process (same tm, mode & type)
Only after all the files are available in GCS, you can move on to step#3.
After all required documents (doc) for the given type data sync (ex: product, order, etc) have been made available in GCS, the DI-SAAS request can be called https://boxalino.atlassian.net/wiki/spaces/BPKB/pages/928874497/DI-SAAS+ELT+Flow#The-DI-request , assigning connector → type : boxalino.
Technical notes
Every request to our nodes returns a response (200 or 400). This can be used internally for testing purposes as well.
It is possible to review the STATUS of the requests at any given time Status Review
0 Comments