开启multiLine：如何在Databricks中按文件读取.flib到Delta表

算力阿峰

2026-7-4

解决方案

方法1：分组拼接文件内容

基于你现有的查询逻辑，通过按文件名分组，将同一文件的所有行内容拼接成完整文本。用collect_list收集当前文件的所有行数据，再通过concat_ws以换行符分隔拼接，还原文件原始结构：

USE CATALOG "data-beta";
USE data_beta_dev;
DROP TABLE IF EXISTS data_beta;
CREATE TABLE data_beta 
USING delta
AS
SELECT
  REVERSE(SUBSTRING_INDEX(REVERSE(_metadata.file_path), '/', 1)) as FileName,
  concat_ws('\n', collect_list(value)) AS FileData
FROM read_files("wasbs://<redacted>@<redacted>.blob.core.windows.net/some/path/to/flibs/*.flib", format => 'text')
GROUP BY REVERSE(SUBSTRING_INDEX(REVERSE(_metadata.file_path), '/', 1));

SELECT * FROM data_beta;

方法2：以二进制文件格式读取（更高效）

直接使用binaryFile格式读取文件，每个文件会被作为单条记录返回，再将二进制内容转换为字符串即可得到完整文件内容：

USE CATALOG "data-beta";
USE data_beta_dev;
DROP TABLE IF EXISTS data_beta;
CREATE TABLE data_beta 
USING delta
AS
SELECT
  REVERSE(SUBSTRING_INDEX(REVERSE(_metadata.file_path), '/', 1)) as FileName,
  CAST(content AS STRING) AS FileData
FROM read_files("wasbs://<redacted>@<redacted>.blob.core.windows.net/some/path/to/flibs/*.flib", format => 'binaryFile');

SELECT * FROM data_beta;