I'm adding a smallish DB to version control (for unit testing purposes). The binary .db is 52 KB. The ascii .dump is 64 KB. The manually edited dump to take advantage of insert into t values (v1),(v2),(v3),...; is only 37 KB.
The compressed sizes are much closer of course, since all the duplicated INSERT INTO "SomeTable" VALUES( fragments compress really well (gzip-9), yet it's still ~5% smaller in this case. Here an uncompressed text form is preferred for diffs in the VCS. Such a dump is close in size of a .csv "pseudo-dump", but more convenient since multi-table and with full fidelity of the schema. Could we please enhance the shell to support this? Thanks, --DD PS: Of course, because of the statement max-length limit, once in a while one must close a very long insert statement, which is very hard to gauge and test for "manually". 37536 Apr 11 10:15 dump-lite.sql 64986 Apr 11 10:11 dump.sql 5994 Apr 11 10:15 dump-lite.sql.gz 6323 Apr 11 10:11 dump.sql.gz
