How do I dump the data, and only the data, not the schema, of some SQLite3 tables of a database (not all the tables)?
The dump should be in SQL format, as it should be easily re-entered into the database later and should be done from the command line. Something like

sqlite3 db .dump

but without dumping the schema and selecting which tables to dump.

You’re not saying what you wish to do with the dumped file.

To get a CSV file (which can be imported into almost everything)

.mode csv 
-- use '.separator SOME_STRING' for something other than a comma.
.headers on 
.out file.csv 
select * from MyTable;

To get an SQL file (which can be reinserted into a different SQLite database)

.mode insert <target_table_name>
.out file.sql 
select * from MyTable;

You can do this getting difference of .schema and .dump commands. for example with grep:

sqlite3 some.db .schema > schema.sql
sqlite3 some.db .dump > dump.sql
grep -vx -f schema.sql dump.sql > data.sql

data.sql file will contain only data without schema, something like this:

BEGIN TRANSACTION;
INSERT INTO "table1" VALUES ...;
...
INSERT INTO "table2" VALUES ...;
...
COMMIT;

I hope this helps you.

You can specify one or more table arguments to the .dump special command, e.g.sqlite3 db ".dump 'table1' 'table2'".

Not the best way, but at lease does not need external tools (except grep, which is standard on *nix boxes anyway)

sqlite3 database.db3 .dump | grep '^INSERT INTO "tablename"'

but you do need to do this command for each table you are looking for though.

Note that this does not include schema.

Any answer which suggests using grep to exclude the CREATE lines or just grab the INSERT lines from the sqlite3 $DB .dump output will fail badly. The CREATE TABLE commands list one column per line (so excluding CREATE won’t get all of it), and values on the INSERT lines can have embedded newlines (so you can’t grab just the INSERT lines).

for t in $(sqlite3 $DB .tables); do
    echo -e ".mode insert $tnselect * from $t;"
done | sqlite3 $DB > backup.sql

Tested on sqlite3 version 3.6.20.

If you want to exclude certain tables you can filter them with $(sqlite $DB .tables | grep -v -e one -e two -e three), or if you want to get a specific subset replace that with one two three.

As an improvement to Paul Egan’s answer, this can be accomplished as follows:

sqlite3 database.db3 '.dump "table1" "table2"' | grep '^INSERT'

–or–

sqlite3 database.db3 '.dump "table1" "table2"' | grep -v '^CREATE'

The caveat, of course, is that you have to have grep installed.

In Python or Java or any high level language the .dump does not work. We need to code the conversion to CSV by hand. I give an Python example. Others, examples would be appreciated:

from os import path   
import csv 

def convert_to_csv(directory, db_name):
    conn = sqlite3.connect(path.join(directory, db_name + '.db'))
    cursor = conn.cursor()
    cursor.execute("SELECT name FROM sqlite_master WHERE type='table';")
    tables = cursor.fetchall()
    for table in tables:
        table = table[0]
        cursor.execute('SELECT * FROM ' + table)
        column_names = [column_name[0] for column_name in cursor.description]
        with open(path.join(directory, table + '.csv'), 'w') as csv_file:
            csv_writer = csv.writer(csv_file)
            csv_writer.writerow(column_names)
            while True:
                try:
                    csv_writer.writerow(cursor.fetchone())
                except csv.Error:
                    break

If you have ‘panel data, in other words many individual entries with id’s add this to the with look and it also dumps summary statistics:

        if 'id' in column_names:
            with open(path.join(directory, table + '_aggregate.csv'), 'w') as csv_file:
                csv_writer = csv.writer(csv_file)
                column_names.remove('id')
                column_names.remove('round')
                sum_string = ','.join('sum(%s)' % item for item in column_names)
                cursor.execute('SELECT round, ' + sum_string +' FROM ' + table + ' GROUP BY round;')
                csv_writer.writerow(['round'] + column_names)
                while True:
                    try:
                        csv_writer.writerow(cursor.fetchone())
                    except csv.Error:
                        break 

Review of other possible solutions

Include only INSERTs

sqlite3 database.db3 .dump | grep '^INSERT INTO "tablename"'

Easy to implement but it will fail if any of your columns include new lines

SQLite insert mode

for t in $(sqlite3 $DB .tables); do
    echo -e ".mode insert $tnselect * from $t;"
done | sqlite3 $DB > backup.sql

This is a nice and customizable solution, but it doesn’t work if your columns have blob objects like ‘Geometry’ type in spatialite

Diff the dump with the schema

sqlite3 some.db .schema > schema.sql
sqlite3 some.db .dump > dump.sql
grep -v -f schema.sql dump > data.sql

Not sure why, but is not working for me

Another (new) possible solution

Probably there is not a best answer to this question, but one that is working for me is grep the inserts taking into account that be new lines in the column values with an expression like this

grep -Pzo "(?s)^INSERT.*);[ t]*$"

To select the tables do be dumped .dump admits a LIKE argument to match the table names, but if this is not enough probably a simple script is better option

TABLES='table1 table2 table3'

echo '' > /tmp/backup.sql
for t in $TABLES ; do
    echo -e ".dump ${t}" | sqlite3 database.db3 | grep -Pzo "(?s)^INSERT.*?);$" >> /tmp/backup.sql
done

or, something more elaborated to respect foreign keys and encapsulate all the dump in only one transaction

TABLES='table1 table2 table3'

echo 'BEGIN TRANSACTION;' > /tmp/backup.sql
echo '' >> /tmp/backup.sql
for t in $TABLES ; do
    echo -e ".dump ${t}" | sqlite3 $1 | grep -Pzo "(?s)^INSERT.*?);$" | grep -v -e 'PRAGMA foreign_keys=OFF;' -e 'BEGIN TRANSACTION;' -e 'COMMIT;' >> /tmp/backup.sql
done

echo '' >> /tmp/backup.sql
echo 'COMMIT;' >> /tmp/backup.sql

Take into account that the grep expression will fail if ); is a string present in any of the columns

To restore it (in a database with the tables already created)

sqlite3 -bail database.db3 < /tmp/backup.sql

According to the SQLite documentation for the Command Line Shell For SQLite you can export an SQLite table (or part of a table) as CSV, simply by setting the “mode” to “csv” and then run a query to extract the desired rows of the table:

sqlite> .header on
sqlite> .mode csv
sqlite> .once c:/work/dataout.csv
sqlite> SELECT * FROM tab1;
sqlite> .exit

Then use the “.import” command to import CSV (comma separated value) data into an SQLite table:

sqlite> .mode csv
sqlite> .import C:/work/dataout.csv tab1
sqlite> .exit

Please read the further documentation about the two cases to consider: (1) Table “tab1” does not previously exist and (2) table “tab1” does already exist.

The best method would be to take the code the sqlite3 db dump would do, excluding schema parts.

Example pseudo code:

SELECT 'INSERT INTO ' || tableName || ' VALUES( ' || 
  {for each value} ' quote(' || value || ')'     (+ commas until final)
|| ')' FROM 'tableName' ORDER BY rowid DESC

See: src/shell.c:838 (for sqlite-3.5.9) for actual code

You might even just take that shell and comment out the schema parts and use that.

The answer by retracile should be the closest one, yet it does not work for my case. One insert query just broke in the middle and the export just stopped. Not sure what is the reason. However It works fine during .dump.

Finally I wrote a tool for the split up the SQL generated from .dump:

https://github.com/motherapp/sqlite_sql_parser/

You could do a select on the tables inserting commas after each field to produce a csv, or use a GUI tool to return all the data and save it to a csv.