The CSV Virtual Table

1. Overview

The CSV virtual table reads RFC 4180 formatted comma-separated values, and returns that content as if it were rows and columns of an SQL table.

The CSV virtual table is useful to applications that need to bulk-load large amounts of comma-separated value content. The CSV virtual table is also useful as a template source file for implementing other virtual tables.

The CVS virtual table is not built into the SQLite amalgamation. It is available as a separate source file that can be compiled into a loadable extension. Typical usage of the CSV virtual table from the command-line shell would be something like this:

.load ./csv
CREATE VIRTUAL TABLE temp.t1 USING csv(filename='thefile.csv');
SELECT * FROM t1;

The first line of the script above causes the command-line shell to read and activate the run-time loadable extension for CSV. For an application, the equivalent C-language API is sqlite3_load_extension(). Observe that the filename extension (ex: ".dll" or ".so" or ".dylib") is omitted from the extension filename. Omitting the filename extension is not required, but it helps in making the script cross-platform. SQLite will automatically append the appropriate extension.

The second line above creates a virtual table named "t1" that reads the content of the file named in the argument. The number and names of the columns is determined automatically by reading the first line of content. Other options to the CSV virtual table provide the ability to take the CSV content from a string rather than a separate file, and give the programmer more control over the number and names of the columns. The options are detailed below. The CSV virtual table is usually created as a TEMP table so that it exists only for the current database connection and does not become a permanent part of the database schema. Note that there is no "CREATE TEMP VIRTUAL TABLE" command in SQLite. Instead, prepend the "temp." schema prefix to the name of the virtual table.

The third line of the example shows the virtual table being used, to read all content of the CSV file. This is perhaps the simplest possible use of the virtual table. The CSV virtual table can be used anywhere an ordinary virtual table can be used. One can use the CSV virtual table inside subqueries, or common table expressions or add WHERE, GROUP BY, HAVING, ORDER BY, and LIMIT clauses as required.

2. Recognized Arguments

The example above showed a single filename='th3file.csv' argument for the CSV virtual table. But other arguments are also possible.

• filename=FILENAME

  The filename= argument specifies an external file from which CSV content is read. Every CSV virtual table must have either a filename= argument or a data= argument and not both.

• data=TEXT

  The data= argument specifies that TEXT is the literal content of the CSV file.

• schema=SCHEMA

  The schema= argument specifies a CREATE TABLE statement that the CVS virtual table passes to the sqlite3_declare_vtab() interface in order to define the number and names of the columns in the virtual table. If both the schema= and the columns= arguments are omitted, then the CVS virtual table reads the first row of the input content in order to determine the number of columns and names the columns cNNN where NNN values are consecutive integers. It is not allowed to have both schema= and columns= arguments.

• columns=N

  The columns=N argument causes the virtual table to have exactly N columns. If the input data contains more columns than this, then the excess columns are ignored. If the input data contains fewer columns, then extra columns are filled with NULL.