Coming from a Big Data (Hadoop / AWS) background, I would say that SQL is a good choice for analyzing data once it is in the data warehouse, but it is not needed to get the data there. Most of the ETL code in this environment is based on Apache Spark 2.x and either Scala or Python as programming language. On a Hortonworks HDP cluster, you would use PySpark directly; on AWS, you would use AWS Glue which allows you to set up and schedule the PySpark code. Although PySpark does not stop you from using SQL, there is a preference to write the transformations in terms of native Spark functions. So, to answer the question: no, SQL is not a primary tool of choice for implementing ETL pipelines in 2019 - PySpark is.