*** DRAFT ***
The SQLite Zipfile Module

1. Overview

The zipfile module provides read/write access to simple ZIP archives. The current implementation has the following restrictions:

Some or all of these restrictions may be removed in the future.

2. Obtaining and Compiling Zipfile

The code for the zipfile module is found in the ext/misc/zipfile.c file of the main SQLite source tree. It may be compiled into an SQLite loadable extension using a command like:

gcc -g -fPIC -shared zipfile.c -o zipfile.so

Alternatively, the zipfile.c file may be compiled into the application. In this case, the following function should be invoked to register the extension with each new database connection:

int sqlite3_zipfile_init(sqlite3 *db, void*, void*);

The first argument passed should be the database handle to register the extension with. The second and third arguments should both be passed 0.

Zipfile is included in most builds of the command-line shell.

3. Using Zipfile

The zipfile module provides three similar interfaces for accessing, updating and creating zip file archives:

  1. A table-valued function, which provides read-only access to existing archives, either from the file-system or in-memory.
  2. A virtual table, which provides read and write access to archives stored in the file-system.
  3. An SQL aggregate function, which can be used to create new archives in memory.

The zipfile module provides two similar interfaces for accessing zip archives. A table-valued function, which provides read-only access to existing archives, and a virtual table interface, which provides both read and write access.

3.1. Table-Valued Function (read-only access)

For reading existing zip archives, the Zipfile module provides a table-valued function that accepts a single argument. If the argument is a text value, then it is a path to a zip archive to read from the file-system. Or, if the argument is an SQL blob, then it is the zip archive data itself.

For example, to inspect the contents of zip archive "test.zip" from the current directory:

SELECT * FROM zipfile('test.zip');

Or, from the SQLite shell tool (the readfile() function reads the contents of a file from the file-system and returns it as a blob):

SELECT * FROM zipfile( readfile('test.zip') );

The table-valued function returns one row for each record (file, directory or symbolic link) in the zip archive. Each row has the following columns:

Column NameContents
name File name/path for the zip file record.
mode UNIX mode, as returned by stat(2) for the zip file record (an integer). This identifies the type of record (file, directory or symbolic link), and the associated user/group/all permissions.
mtime UTC timestamp, in seconds since the UNIX epoch (an integer).
sz Size of associated data in bytes after it has been uncompressed (an integer).
rawdata Raw (possibly compressed) data associated with zip file entry (a blob).
data If the compression method for the record is either 0 or 8 (see below), then the uncompressed data associated with the zip file entry. Or, if the compression method is not 0 or 8, this column contains a NULL value.
method The compression method used to compress the data (an integer). The value 0 indicates that the data is stored in the zip archive without compression. 8 means the raw deflate algorithm.

3.2. Virtual Table Interface (read/write access)

In order to create or modify an existing zip file, a "zipfile" virtual table must be created in the database schema. The CREATE VIRTUAL TABLE statement expects a path to the zip file as its only argument. For example, to write to zip file "test.zip" in the current directory, a zipfile table may be created using:

CREATE VIRTUAL TABLE temp.zip USING zipfile('test.zip');

Such a virtual table has the same columns as the table-valued function described in the previous section. It may be read from using a SELECT statement in the same way as the table-valued function can.

Using the virtual table interface, new entries may be added to a zip archive by inserting new rows into the virtual table. Entries may be removed by deleting rows or modified by updating them.

3.2.1. Adding Entries to a Zip Archive

Entries may be added to a zip archive by inserting new rows. The easiest way to do this is to specify values for the "name" and "data" columns only and have zipfile fill in sensible defaults for other fields. To insert a directory into the archive, set the "data" column to NULL. For example, to add the directory "dir1" and the file "m.txt" containing the text "abcdefghi" to zip archive "test.zip":

INSERT INTO temp.zip(name, data) VALUES('dir1', NULL);           -- Add directory 
INSERT INTO temp.zip(name, data) VALUES('m.txt', 'abcdefghi');   -- Add regular file 

When a directory is inserted, if the "name" value does not end with a '/' character, the zipfile module appends one. This is necessary for compatibility with other programs (most notably "info-zip") that manipulate zip archives.

To insert a symbolic link, the user must also supply a "mode" value. For example, to add a symbolic link from "link.txt" to "m.txt":

INSERT INTO temp.zip(name, mode, data) VALUES('link.txt', 'lrwxrw-rw-', 'abcdefghi');

The following rules and caveats apply to the values specified as part of each INSERT statement:

Columns Notes
name A non-NULL text value must be specified for the name column. It is an error if the specified name already exists in the archive.
mode If NULL is inserted into the mode column, then the mode of the new archive entry is automatically set to either 33188 (-rw-r--r--) or 16877 (drwxr-xr-x), depending on whether or not the values specified for columns "sz", "data" and "rawdata" indicate that the new entry is a directory.

If the specified value is an integer (or text that looks like an integer), it is inserted verbatim. If the value is not a valid UNIX mode, some programs may behave unexpectedly when extracting files from the archive.

Finally, if the value specified for this column is not an integer or a NULL, then it is assumed to be a UNIX permissions string similar to those output by the "ls -l" command (e.g. "-rw-r--r--", "drwxr-xr-x" etc.). In this case, if the string cannot be parsed it is an error.
mtime If NULL is inserted into the mtime column, then the timestamp of the new entry is set to the current time. Otherwise, the specified value is interpreted as an integer and used as is.
sz This column must be set to NULL. If a non-NULL value is inserted into this column, or if a new non-NULL value is provided using an UPDATE statement, it is an error.
rawdata This column must be set to NULL. If a non-NULL value is inserted into this column, or if a new non-NULL value is provided using an UPDATE statement, it is an error.
data To insert a directory into the archive, this field must be set to NULL. In this case if a value was explicitly specified for the "mode" column, then it must be consistent with a directory (i.e. it must be true that (mode & 0040000)=0040000).

Otherwise, the value inserted into this field is the file contents for a regular file, or the target of a symbolic link.
method This field must be set one of integer values 0 and 8, or else to NULL.

For a directory entry, any value inserted into this field is ignored. Otherwise, if it is set to 0, then the file data or symbolic link target is stored as is in the zip archive and the compression method set to 0. If it is set to 8, then the file data or link target is compressed using deflate compression before it is stored and the compression method set to 8. Finally, if a NULL value is written to this field, the zipfile module automatically decides whether or not to compress the data before storing it.

Specifying an explicit value for the rowid field as part of an INSERT statement is not supported. Any value supplied is ignored.

3.2.2. Deleting Zip Archive Entries

Records may be removed from an existing zip archive by deleting the corresponding rows. For example, to remove file "m.txt" from zip archive "test.zip" using the virtual table created above:

DELETE FROM temp.zip WHERE name = 'm.txt';

Note that deleting records from a zip archive does not reclaim the space used within the archive - it merely removes an entry from the archives "Central Directory Structure", making the entry inaccessible. One way to work around this inefficiency is to create a new zip archive based on the contents of the edited archive. For example, after editing the archive accessed via virtual table temp.zzz:

-- Create a new, empty, archive: 
CREATE VIRTUAL TABLE temp.newzip USING zipfile('new.zip');

-- Copy the contents of the existing archive into the new archive
INSERT INTO temp.newzip(name, mode, mtime, data, method)
    SELECT name, mode, mtime, data, method FROM temp.zzz;

3.2.3. Updating Existing Zip Archive Entries

Existing zip archive entries may be modified using UPDATE statements.

The three leftmost columns of a zipfile virtual table, "name", "mode" and "mtime", may each be set to any value that may be inserted into the same column (see above). If either "mode" or "mtime" is set to NULL, the final value is determined as described for an INSERT of a NULL value - the current time for "mtime" and either 33188 or 16877 for "mode", depending on whether or not the values specified for the next four columns of the zipfile table indicate that the entry is a directory or a file.

It is an error to attempt to set the sz or rawdata field to any value other than NULL.

The data and method columns may also be set as described for an INSERT above.

3.3. The zipfile() Aggregate Function

New zip archives may be constructed entirely within memory using the zipfile() aggregate function. Each row visited by the aggregate function adds an entry to the zip archive. The value returned is a blob containing the entire archive image.

The zipfile() aggregate function may be called with 2, 4 or 5 arguments. If it is called with 5 arguments, then the entry added to the archive is equivalent to inserting the same values into the "name", "mode", "mtime", "data" and "method" columns of a zipfile virtual table.

If zipfile() is invoked with 2 arguments, then the entry added to the archive is equivalent to that added by inserting the same two values into the "name" and "data" columns of a zipfile virtual table, with all other values set to NULL. If invoked with 4 arguments, it is equivalent to inserting the 4 values into the "name", "mode", "mtime" and "data" columns. In other words, the following pairs of queries are equivalent:

SELECT zipfile(name, data) ...
SELECT zipfile(name, NULL, NULL, data, NULL) ...

SELECT zipfile(name, mode, mtime, data) ...
SELECT zipfile(name, mode, mtime, data, NULL) ...

For example, to create an archive containing two text files, "a.txt" and "b.txt", containing the text "abc" and "123" respectively:

WITH contents(name, data) AS (
  VALUES('a.txt', 'abc') UNION ALL
  VALUES('b.txt', '123')
SELECT zipfile(name, data) FROM contents;

*** DRAFT ***