mariadb-import

mariadb-import defaults to TAB-separated fields regardless of file extension, so for real CSV pass --fields-terminated-by=, and --fields-optionally-enclosed-by='"'. Add --ignore-lines=1 to skip a header row and --local so the client reads the file.

There is no table argument; the table name comes from the data file's base name minus directory and extension, so users.csv loads into table users. The first non-option argument is the database.

--replace overwrites existing rows on duplicate keys while --ignore keeps the existing row, and the two flags are mutually exclusive. Combining them causes the tool to error and exit.

The --lock-tables flag disables parallelism, so the threaded path from --parallel or --use-threads runs only when locking is off. Remove --lock-tables to get concurrent file loads.

Without --local, the server reads the file, requiring the FILE privilege and secure_file_priv compliance. With --local or -L, the client reads it via LOAD DATA LOCAL INFILE, subject to local_infile being enabled.

The default connection character set is auto-detected from the client locale, not a fixed latin1. Pass --default-character-set=utf8mb4 to make the encoding explicit.