将字符串列表转换为表格

Question

I'm just a beginner in sql and I would like to transform strings into a table

Date of application: 01/02/2018 Request: Buy books Contact: email: hi@gmail.com Tel: 0123456789 Ordered inquiry: Order ID: 12345678 BL: 87654321 Product: 123456 Books

Date of application: 01/04/2018 Request: Retour table Contact: Rodion Raskólnikov email: hello@outlook.com Tel: 9876543210 Ordered inquiry: Order Id: 87654321 BL: 12345678 Product: 654321 Tables

Like this:

I tried this:

WITH raw_messages AS (SELECT lines
  FROM `my_table` 
  WHERE REGEXP_CONTAINS(lines, '^Date of application: '))

SELECT 
  REGEXP_EXTRACT(lines, r'^Date of application: [0-9]{2}/[0-9]{2}/[0-9]{4}') as date

FROM raw_messages

It does not work as I would like and I have no idea how to continue building my table.

Answer 1

Below is for BigQuery Standard SQL

Assuming order of fields in your lines are set as it is in your example

#standardSQL
WITH raw_messages AS (
  SELECT lines
  FROM `my_table` 
  WHERE REGEXP_CONTAINS(lines, '^Date of application: ')
)
SELECT 
  REGEXP_EXTRACT(lines, r'(?i)^Date of application: ([0-9]{2}/[0-9]{2}/[0-9]{4})') AS DATE,
  REGEXP_EXTRACT(lines, r'(?i) Request: (.*?) Contact: ') AS request,
  REGEXP_EXTRACT(lines, r'(?i) Contact: (.*?) email: ') AS contact,
  REGEXP_EXTRACT(lines, r'(?i) email: (.*?) Tel: ') AS email,
  REGEXP_EXTRACT(lines, r'(?i) Tel: (.*?) Ordered inquiry: ') AS phone,
  REGEXP_EXTRACT(lines, r'(?i) Order ID: (.*?) BL: ') AS id,
  REGEXP_EXTRACT(lines, r'(?i) BL: (.*?) Product: ') AS bl,
  REGEXP_EXTRACT(lines, r'(?i) Product: (.*?)$') AS product
FROM raw_messages   

You can test , play with above using dummy data from your question as below

#standardSQL
WITH `project.dataset.my_table` AS (
  SELECT 'Date of application: 01/02/2018 Request: Buy books Contact: email: hi@gmail.com Tel: 0123456789 Ordered inquiry: Order ID: 12345678 BL: 87654321 Product: 123456 Books' lines UNION ALL
  SELECT 'Date of application: 01/04/2018 Request: Retour table Contact: Rodion Raskólnikov email: hello@outlook.com Tel: 9876543210 Ordered inquiry: Order Id: 87654321 BL: 12345678 Product: 654321 Tables'
), raw_messages AS (
  SELECT lines
  FROM `project.dataset.my_table` 
  WHERE REGEXP_CONTAINS(lines, '^Date of application: ')
)
SELECT 
  REGEXP_EXTRACT(lines, r'(?i)^Date of application: ([0-9]{2}/[0-9]{2}/[0-9]{4})') AS DATE,
  REGEXP_EXTRACT(lines, r'(?i) Request: (.*?) Contact: ') AS request,
  REGEXP_EXTRACT(lines, r'(?i) Contact: (.*?) email: ') AS contact,
  REGEXP_EXTRACT(lines, r'(?i) email: (.*?) Tel: ') AS email,
  REGEXP_EXTRACT(lines, r'(?i) Tel: (.*?) Ordered inquiry: ') AS phone,
  REGEXP_EXTRACT(lines, r'(?i) Order ID: (.*?) BL: ') AS id,
  REGEXP_EXTRACT(lines, r'(?i) BL: (.*?) Product: ') AS bl,
  REGEXP_EXTRACT(lines, r'(?i) Product: (.*?)$') AS product
FROM raw_messages    

with result

Row DATE        request         contact             email               phone       id          bl          product  
1   01/02/2018  Buy books       null                hi@gmail.com        0123456789  12345678    87654321    123456 Books     
2   01/04/2018  Retour table    Rodion Raskólnikov  hello@outlook.com   9876543210  87654321    12345678    654321 Tables    

Answer 2

In case if order of fields in your strings is not known/guaranteed, but you know all fields in them - below is smart enough to parse those properly

#standardSQL
WITH raw_messages AS (
  SELECT lines FROM `project.dataset.my_table` 
  WHERE REGEXP_CONTAINS(lines, '^Date of application: ')
), fields AS (
  SELECT 'Date of application' field, 'date' column UNION ALL
  SELECT 'Request', 'request' UNION ALL
  SELECT 'Contact', 'contact' UNION ALL
  SELECT 'email', 'email' UNION ALL
  SELECT 'Tel', 'phone' UNION ALL
  SELECT 'Order ID', 'id' UNION ALL
  SELECT 'BL', 'bl' UNION ALL
  SELECT 'Product', 'product' UNION ALL
  SELECT 'Ordered inquiry', '' UNION ALL
  SELECT 'Boundary of string', ''
), patterns AS ( 
  SELECT f1.field, f1. column, CONCAT(r'(?i) ',f1.field,': (.*)',f2.field,': ') pattern
  FROM fields f1 CROSS JOIN fields f2
), splits AS (SELECT ARRAY(
      SELECT AS STRUCT column, ARRAY_AGG(value ORDER BY LENGTH(value) LIMIT 1)[OFFSET(0)] value
      FROM (SELECT column, REGEXP_EXTRACT(CONCAT(' Boundary of string: ', lines, ' Boundary of string: '), pattern) value
        FROM patterns ) 
      WHERE NOT value IS NULL AND NOT column = '' GROUP BY column 
    ) arr FROM raw_messages
) SELECT 
  (SELECT value FROM UNNEST(arr) WHERE column='date')     AS DATE,
  (SELECT value FROM UNNEST(arr) WHERE column='request')  AS request,
  (SELECT value FROM UNNEST(arr) WHERE column='contact')  AS contact,
  (SELECT value FROM UNNEST(arr) WHERE column='email')    AS email,
  (SELECT value FROM UNNEST(arr) WHERE column='phone')    AS phone,
  (SELECT value FROM UNNEST(arr) WHERE column='id')       AS id,
  (SELECT value FROM UNNEST(arr) WHERE column='bl')       AS bl,
  (SELECT value FROM UNNEST(arr) WHERE column='product')  AS product
FROM splits     

you can test above with the same dummy data as in my another answer and obviously result should be the same

Note: as you can see - you need to set fields AS (...) CTE explicitly with all fields in strings and respective column names to be used in any order but important - you need to add one more entry there - 'Boundary of string'