Sql loader for sqlite. [csv2sqlite, csv to sqlite, CSV to SQLite]
C++ C Shell Python Makefile
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Sqlite Loader

Imports csv files to sqlite database file


During one of my personal projects, I had to use '.import' command of the sqlite3 shell. The .import command is very limited in functionality. It also always binds column data as text. This has performance hit as well as disk usage hit. I also need some very basic conversion support like store numbers as numbers in db files as well as some way to parse date related data.

All these needs leads for the development of this project. Apart from the above mentioned features, one of the key requirement for this project is performance.


  • Highly performant by writing it in C/C++. It also used sqlite features to optimize for speed.
  • Support triming text
  • Support date/time input formats
  • Support pivotYear for date/time
  • Support missing value declaration for columns
  • Separate types for int, real and text as opposed to always text with '.import' option of sqlite3 shell
  • Support skipping of input fields in the csv file
  • Support storing dates as text or epoch

Usage for project

This project is available as a command line utility which can be used to create sqltie db files from csv files.

TBD - Fixed length file support might come in future. Right now CSV looks like a better choice


  -h, --help            show this help message and exit
  --version             show program's version number and exit
  -t <table_name>       Table to be created
  -i <input_file>       Input fixed length text file containing records to be
                        ported to sqlite data base.
  -o <output_file>      Output file name for the sqlite data base.
  -l <layout_file>      Layout file containing field definitions. These
                                   field definitions will be used for parsing
                        fixed length record
  -c <N>                Number of records after which commit operation is called
                        on the data base
  -d                    Delete if table already exists
  -a                    Append to existing table if it does exist
  -v                    Debug mode
  -s                    Show stats
  -p <semi-colon-separated-pragma-list>
                        Semi colon separated pragma which ran before creation of DB.
                         Example: -p "pragma page_size=4096;pragma cache_size=1000;"
                        If you specify page_size pragma then it will override
                        the pageSize specifield in the layout file. Refer detail
                        help with -h
  -b <N>                Read buffer size in bytes. Default: 1048576
  -f <N>                Field buffer size in bytes. Default: 1048576

Layout Definition

General layout structure is define in json format:


Layout Definition Parameters

Parameter Name Description
name Layout name that will be used as table name. Can also be passed as an argument to utility.
type csv. 'csv' for delimited file.
separator Separator used for file. Only valid for csv files.
storeDateAsEpoch Store date as Epoch seconds. This will help reduce file size.
isRowId Controls creation of internal row id column in sqlite db file. Default is true. Refer for details https://www.sqlite.org/withoutrowid.html
pageSize Controls page size for output db file. This way you can specify different page size for different files. For more details refer to https://www.sqlite.org/pragma.html#pragma_page_size
fieldList Array of field definitions
primaryKey Primary key definitions
indexList Index definitions array

Layout Field Definition Parameters

Parameter Name Description
name Field name. It will become the column name in db
type text/integer/real/date/time. 'text' text field like "hello". 'integer' for integers like 10. 'real' for decimals like 5.6
format If type is date/time then we need to provide this. Use date format from http://pubs.opengroup.org/onlinepubs/009695399/functions/strftime.html. Example: %m%d%Y for 11022014 which is 2014-02-11. %d/%m/%Y-%H-%M-%S for 02/11/2014-21-56-53 which is 2014-02-11T21:56:53
pivotYear If type is date/time then we can provide pivot year. Refer to http://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormatter.html#withPivotYear(int). Example: 2000
missingValue If this is the value then it will be replaced with null in db.
isSkip If true, this field is skipped while parsing the input file. Default is false
isTrim If type is text, then if false, this field is NOT trimmed. Default is true

Primary Key Definition Parameters

Parameter Name Description
columnList Column list of all column names of primary key

Primary Key Column List Definition Parameters

Parameter Name Description
name Column name

Index Definition Parameters

Parameter Name Description
name Name for the index. If left blank it will be auto generated. Example: idx_col_name_1
isUnique Creates unique index if true. Default is false
whereClause Where clause for creating partial index. "WHERE" will be appended automatically.
columnList Index column definition array. Example: idx_col_name_1

Index Column Definition Parameters

Parameter Name Description
name Column name on which index is based on.

Example of csv layout:

    "name" : "in",
    "type" : "csv",
    "separator" : "   ",
    "isRowId" : "true",
    "pageSize": 65536,
    "primaryKey" : {
      "columnList" : [
            "name" : "record_id"
    "indexList": [
            "columnList": [
                    "name": "name"
                    "name": "balance"
    "fieldList": [
            "name": "record_id",
            "type": "integer"
            "name": "name"
            "name": "balance",
            "type": "real"
            "name": "date1",
            "type": "date",
            "format": "%m%d%Y"
            "name": "date2",
            "type": "date",
            "format": "%Y-%m-%d"
            "name": "time1",
            "type": "time",
            "format": "%Y-%m-%dT%H:%M:%S"
            "name": "time2",
            "type": "time",
            "format": "%d/%m/%Y-%H-%M-%S"