cqlsh - inconsistent escaping of non-printable characters

[nyh]

This issue in cqlsh caused @juliayakovlev and me a lot of grief. It was also reported in Cassandra years ago - see https://issues.apache.org/jira/browse/CASSANDRA-8790 - but never solved.

The problem is that cqlsh escapes non-printable characters in text columns in a certain way, and then doesn't allow you to use the same format in queries. For example:

CREATE KEYSPACE ks WITH REPLICATION = { 'class' : 'SimpleStrategy', 'replication_factor': 1 };
USE ks;
CREATE TABLE cf (pk text primary key);
INSERT INTO cf (pk) values ('^A');     # "^A" is control-A. Type control-V and then control

When we use SELECT we see the following:

 pk
------
 \x01
(1 rows)

Note how the character 1 (control-A) is shown here as '\x01'.

However, one cannot use this syntax on input: The string '\x01' is treated literally as those 4 characters, and not translated to character 1:

cqlsh:ks> select * from cf where pk='\x01';
 pk
----
(0 rows)

To read this row, one needs to type control-A (usually control-V control-A) again.. But there's no way to guess this from the SELECT output.

cqlsh:ks> select * from cf where pk='^A';
 pk
------
 \x01
(1 rows)