SWI-Prolog C-library


Jan Wielemaker
HCS,
University of Amsterdam
The Netherlands
E-mail: wielemak@science.uva.nl

Abstract

This document describes commonly used foreign language extensions to SWI-Prolog distributed as a package known under the name clib. The package defines a number of Prolog libraries with accompagnying foreign libraries.

library(unix)
This library provides Unix process control using fork(), exec(), pipe(), etc.
library(files)
This library provides low-level access to file objects.
library(cgi)
This library provides access to CGI form-data if Prolog is used for CGI-scripting.
library(crypt)
This library provides access to Unix password encryption.
library(mime)
This library decodes MIME messages.
library(socket)
This library provides access to TCP/IP communication.
library(streampool)
Dispatch input from a pool of (socket) streams.
library(time)
This library provides timing (alarm) functions.
library(rlimit)
Provides access to POSIX resource limits (CPU, memory, files).

On Windows systems, the library(unix) and library(crypt) libraries can only be used if the whole SWI-Prolog suite is compiled using Cywin. The other libraries have been ported to native Windows.

Table of Contents

1 Introduction

Many useful facilities offered by one or more of the operating systems supported by SWI-Prolog are not supported by the SWI-Prolog kernel distribution. Including these would enlarge the footprint and complicate portability matters while supporting only a limited part of the user-community.

This document describes library(unix) to deal with the Unix process API, library(socket) to deal with inet-domain stream-sockets, library(cgi) to deal with getting CGI form-data if SWI-Prolog is used as a CGI scripting language and library(crypt) to provide access to Unix password encryption.

2 Unix Process manipulation library

The library(unix) library provides the commonly used Unix primitives to deal with process management. These primitives are useful for many tasks, including server management, parallel computation, exploiting and controlling other processes, etc.

The predicates are modelled closely after their native Unix counterparts. Higher-level primitives, especially to make this library portable to non-Unix systems are desirable. Using these primitives and considering that process manipulation is not a very time-critical operation we anticipate these libraries to be developed in Prolog.

fork(-Pid)
Clone the current process into two branches. In the child, Pid is unified to child. In the original process, Pid is unified to the process identifier of the created child. Both parent and child are fully functional Prolog processes running the same program. The processes share open I/O streams that refer to Unix native streams, such as files, sockets and pipes. Data is not shared, though on most Unix systems data is initially shared and duplicated only if one of the programs attempts to modify the data.

Unix fork() is the only way to create new processes and fork/2 is a simple direct interface to it.

exec(+Command(...Args...))
Replace the running program by starting Command using the given commandline arguments. Each command-line argument must be atomic and is converted to a string before passed to the Unix call execvp().

Unix exec() is the only way to start an executable file executing. It is commonly used together with fork/1. For example to start netscape on an URL in the background, do:


run_netscape(URL) :-
        (    fork(child),
             exec(netscape(URL))
        ;    true
        ).

Using this code, netscape remains part of the process-group of the invoking Prolog process and Prolog does not wait for netscape to terminate. The predicate wait/2 allows waiting for a child, while detach_IO/0 disconnects the child as a deamon process.

wait(-Pid, -Status)
Wait for a child to change status. Then report the child that changed status as well as the reason. Status is unified with exited(ExitCode) if the child with pid Pid was terminated by calling exit() (Prolog halt/[0,1]). ExitCode is the return=status. Status is unified with signaled(Signal) if the child died due to a software interrupt (see kill/2). Signal contains the signal number. Finally, if the process suspended execution due to a signal, Status is unified with stopped(Signal).

kill(+Pid, +Signal)
Deliver a software interrupt to the process with identifier Pid using software-interrupt number Signal. See also on_signal/2. The meaning of the signal numbers can be found in the Unix manual. (1).

pipe(-InSream, -OutStream)
Create a communication-pipe. This is normally used to make a child communicate to its parent. After pipe/2, the process is cloned and, depending on the desired direction, both processes close the end of the pipe they do not use. Then they use the remaining stream to communicate. Here is a simple example:


:- use_module(library(unix)).

fork_demo(Result) :-
        pipe(Read, Write),
        fork(Pid),
        (   Pid == child
        ->  close(Read),
            format(Write, '~q.~n',
                   [hello(world)]),
            flush_output(Write),
            halt
        ;   close(Write),
            read(Read, Result),
            close(Read)
        ).

dup(+FromStream, +ToStream)
Interface to Unix dup2(), copying the underlying filedescriptor and thus making both streams point to the same underlying object. This is normally used together with fork/1 and pipe/2 to talk to an external program that is designed to communicate using standard I/O.

Both FromStream and ToStream either refer to a Prolog stream or an integer descriptor number to refer directly to OS descriptors. See also demo/pipe.pl in the source-distribution of this package.

detach_IO
This predicate is intended to create Unix deamon-processes. It preforms two actions. First of all, the I/O streams user_input, user_output and user_error are closed and rebound to a Prolog stream that returns end-of-file on any attempt to read and starts writing to a file named /tmp/pl-out.pid (where <pid> is the process-id of the calling Prolog) on any attempt to write. This file is opened only if there is data available. This is intended for debugging purposes. (2) Finally, the process is detached from the current process-group and its controlling terminal.

3 File manipulation library

The library(files) library provides additional operations on files from SWI-Prolog. It is currently very incomplete.

set_time_file(+File, -OldTimes, +NewTimes)
Query and set POSIX time attributes of a file. Both OldTimes and NewTimes are lists of option-terms. Times are represented in SWI-Prolog's standard floating point numbers. New times may be specified as now to indicate the current time. Defined options are:

access(Time)
Describes the time of last access of the file. This value can be read and written.

modified(Time)
Describes the time the contents of the file was last modified. This value can be read and written.

changed(Time)
Describes the time the file-structure itself was changed by adding (link()) or removing (unlink()) names.

Here are some example queries. The first retrieves the access-time, while the second sets the last-modified time to the current time.


?- set_time_file(foo, [acess(Access)], []).
?- set_time_file(foo, [], [modified(now)]).

4 Socket library

The library(socket) library provides TCP inet-domain sockets from SWI-Prolog, both client and server-side communication. The interface of this library is very close to the Unix socket interface, also supported by the MS-Windows winsock API. Since SWI-Prolog 4.0, XPCE is part of SWI-Prolog and offers socket. XPCE provides an event-driven interface to sockets, handling multiple open sockets in paralel.

In the future we hope to provide a more high-level socket interface defined in Prolog and based on these primitives.

tcp_socket(-SocketId)
Creates an INET-domain stream-socket and unifies an identifier to it with SocketId. On MS-Windows, if the socket library is not yet initialised, this will also initialise the library.

tcp_close_socket(+SocketId)
Closes the indicated socket, making SocketId invalid. Normally, sockets are closed by closing both stream handles returned by open_socket/3. There are two cases where tcp_close_socket/1 is used because there are no stream-handles:

tcp_open_socket(+SocketId, -InStream, -OutStream)
Open two SWI-Prolog I/O-streams, one to deal with input from the socket and one with output to the socket. If tcp_bind/2 has been called on the socket. OutSream is useless and will not be created. After closing both InStream and OutSream, the socket itself is discarded.

tcp_bind(+Socket, ?Port)
Bind the socket to Port on the current machine. This operation, together with tcp_listen/2 and tcp_accept/3 implement the server-side of the socket interface. If Port is unbound, the system picks an arbitrary free port and unifies Port with the selected port number.

tcp_listen(+Socket, +Backlog)
Tells, after tcp_bind/2, the socket to listen for incoming requests for connections. Backlog indicates how many pending connection requests are allowed. Pending requests are requests that are not yet acknowledged using tcp_accept/3. If the indicated number is exceeded, the requesting client will be signalled that the service is currently not available. A suggested default value is 5.

tcp_accept(+Socket, -Slave, -Peer)
This predicate waits on a server socket for a connection request by a client. On success, it creates a new socket for the client and binds the identifier to Slave. Peer is bound to the IP-address of the client.

tcp_connect(+Socket, +Host:+Port)
Client-interface to connect a socket to a given Port on a given Host. After successful completion, tcp_open_socket/3 can be used to create I/O-Streams to the remote socket.

tcp_setopt(+Socket, +Option)
Set options on the socket. Defined options are:

reuseaddr
Allow servers to reuse a port without the system being completely sure the port is no longer in use.

dispatch(Bool)
In GUI environments (using XPCE or the Windows plwin.exe executable) this flags defines whether or not any events are dispatched on behalf of the user interface. Default is true. Only very specific situations require setting this to false.

tcp_fcntl(+Stream, +Action, ?Argument)
Interface to the Unix fcntl() call. Currently only suitable to deal switch stream to non-blocking mode using:


        ...
        tcp_fcntl(Stream, setfl. nonblock),
        ...

As of SWI-Prolog 3.2.4, handling of non-blocking stream is supported. An attempt to read from a non-blocking stream returns -1 (or end_of_file for read/1), but at_end_of_stream/1 fails. On actual end-of-input, at_end_of_stream/1 succeeds.

tcp_host_to_address(?HostName, ?Address)
Translate between a machines host-name and it's (IP-)address. If HostName is an atom, it is resolved using gethostbyname() and the IP-number is unified to Address using a term of the format ip(Byte1, Byte2, Byte3, Byte4). Otherwise, if Address is bound to a ip/4 term, it is resolved by gethostbyaddr() and the canonical hostname is unified with HostName.

gethostname(-Hostname)
Return the official fully qualified name of this host. This is achieved by calling gethostname() followed by gethostbyname() and return the official name of the host (h_name) of the structure returned by the latter function.

4.1 Server applications

The typical sequence for generating a server application is defined below:


create_server(Port) :-
        tcp_socket(Socket),
        tcp_bind(Socket, Port),
        tcp_listen(Socket, 5),
        tcp_open_socket(Socket, AcceptFd, _),
        <dispatch>

There are various options for <dispatch>. One is to keep track of active clients and server-sockets using wait_for_input/3. If input arrives at a server socket, use tcp_accept/3 and add the new connection to the active clients. Otherwise deal with the input from the client. Another is to use (Unix) fork/1 to deal with the client in a separate process.

Using fork/1, <dispatch> may be implemented as:


dispatch(AcceptFd) :-
        tcp_accept(AcceptFd, Socket, _Peer),
        fork(Pid)
        (   Pid == child
        ->  tcp_open_socket(Socket, In, Out),
            handle_service(In, Out),
            close(In),
            close(Out),
            halt
        ;   tcp_close_socket(Socket)
        ),
        dispatch(AcceptFd).

4.2 Client applications

The skeleton for client-communication is given below.


create_client(Host, Port) :-
        tcp_socket(Socket),
        tcp_connect(Socket, Host:Port),
        tcp_open_socket(Socket, ReadFd, WriteFd),
        <handle I/O using the two streams>
        close(ReadFd),
        close(WriteFd).

To deal with timeouts and multiple connections, wait_for_input/3 and/or non-blocking streams (see tcp_fcntl/3) can be used.

4.3 The stream_pool library

The library(streampool) library dispatches input from multiple streams based on wait_for_input/3. It is part of the clib package as it is used most of the time together with the library(socket) library. On non-Unix systems it often can only be used with socket streams.

With SWI-Prolog 5.1.x, multi-threading often provides a good alternative to using this library. In this schema one thread watches the listening socket waiting for connections and either creates a thread per connection or processes the accepted connections with a pool of worker threads. The library library(http/thread_httpd) provides an example realising a mult-threaded HTTP server.

add_stream_to_pool(+Stream, :Goal)
Add Stream, which must be an input stream and ---on non-unix systems--- connected to a socket to the pool. If input is available on Stream, Goal is called.

delete_stream_from_pool(+Stream)
Delete the given stream from the pool. Succeeds, even if Stream is no member of the pool. If Stream is unbound the entire pool is emtied but unlike close_stream_pool/0 the streams are not closed.

close_stream_pool
Empty the pool, closing all streams that are part of it.

dispatch_stream_pool(+TimeOut)
Wait for maximum of TimeOut for input on any of the streams in the pool. If there is input, call the Goal associated with add_stream_to_pool/2. If Goal fails or raises an exception a message is printed. TimeOut is described with wait_for_input/3.

If Goal is called, there is some input on the associated stream. Goal must be careful not to block as this will block the entire pool. (3)

stream_pool_main_loop
Calls dispatch_stream_pool/1 in a loop until the pool is empty.

Below is a very simple example that reads the first line of input and echos it back.


:- use_module(library(streampool)).

server(Port) :-
        tcp_socket(Socket),
        tcp_bind(Socket, Port),
        tcp_listen(Socket, 5),
        tcp_open_socket(Socket, In, _Out),
        add_stream_to_pool(In, accept(Socket)),
        stream_pool_main_loop.

accept(Socket) :-
        tcp_accept(Socket, Slave, Peer),
        tcp_open_socket(Slave, In, Out),
        add_stream_to_pool(In, client(In, Out, Peer)).

client(In, Out, _Peer) :-
        read_line_to_codes(In, Command),
        close(In),
        format(Out, 'Please to meet you: ~s~n', [Command]),
        close(Out),
        delete_stream_from_pool(In).

5 CGI Support library

This is currently a very simple library, providing support for obtaining the form-data for a CGI script:

cgi_get_form(-Form)
Decodes standard input and the environment variables to obtain a list of arguments passed to the CGI script. This predicate both deals with the CGI GET method as well as the POST method. If the data cannot be obtained, an existence_error exception is raised.

Below is a very simple CGI script that prints the passed parameters. To test it, compile this program using the command below, copy it to your cgi-bin directory (or make it otherwise known as a CGI-script) and make the query http://myhost.mydomain/cgi-bin/cgidemo?hello=world


% pl -o cgidemo --goal=main --toplevel=halt -c cgidemo.pl


:- use_module(library(cgi)).

main :-
        cgi_get_form(Arguments),
        format('Content-type: text/html~n~n', []),
        format('<HTML>~n', []),
        format('<HEAD>~n', []),
        format('<TITLE>Simple SWI-Prolog CGI script</TITLE>~n', []),
        format('</HEAD>~n~n', []),
        format('<BODY>~n', []),
        format('<P>', []),
        print_args(Arguments),
        format('</BODY>~n</HTML>~n', []).

print_args([]).
print_args([A0|T]) :-
        A0 =.. [Name, Value],
        format('<B>~w</B>=<EM>~w</EM><BR>~n', [Name, Value]),
        print_args(T).

5.1 Some considerations

Printing an HTML document using format/2 is not really a neat way of producing HTML. A high-level alternative is provided by library(http/html_write) from the XPCE package.

6 MIME decoding library

MIME (Multipurpose Internet Mail Extensions) is a format for serializing multiple typed data objects. It was designed for E-mail, but it is also used for other applications such packaging multiple values using the HTTP POST request on web-servers. Double Precision, Inc. has produced the C-libraries rfc822 (mail) and rfc2045 (MIME) for decoding and manipulating MIME messages. The library(mime) library is a Prolog wrapper around the rfc2045 library for decoding MIME messages.

The general name `mime' is used for this library as it is anticipated to add MIME-creation functionality to this library.

Currently the mime library defines one predicate:

mime_parse(Data, Parsed)
Parse Data and unify the result to Parsed. Data is one of:

stream(Stream)
Parse the data from Stream upto the end-of-file.

stream(Stream, Length)
Parse a maximum of Length characters from Stream or upto the end-of-file, whichever comes first.

Text
Atoms, strings, code- and character lists are treated as valid sources of data.

Parsed is a tree structure of mime(Attributes, Data, PartList) terms. Currently either Data is the empty atom or PartList is an empty list. (4) Data is an atom holding the message data. The library automatically decodes base64 and quoted-printable messages. See also the transfer_encoding attribute below.

PartList is a list of mime/3 terms. Attributes is a list holding a subset of the following arguments. For details please consult the RFC2045 document.

type(Atom)
Denotes the Content-Type, how the Data should be interpreted.

transfer_encoding(Atom)
How the Data was encoded. This is not very interesting as the library decodes the content of the message.

character_set(Atom)
The character set used for text data. Note that SWI-Prolog's capabilities for character-set handling are limited.

language(Atom)
Language in which the text-data is written.

id(Atom)
Identifier of the message-part.

description(Atom)
Descrptive text for the Data.

disposition(Atom)
Where the data comes from. The current library only deals with `inline' data.

name(Atom)
Name of the part.

filename(Atom)
Name of the file the data should be stored in.

7 Unix password encryption library

The library(crypt) library defines crypt/2 for encrypting and testing Unix passwords:

crypt(+Plain, ?Encrypted)
This predicate can be used in three modes. If Encrypted is unbound, it will be unified to a string (list of character-codes) holding a random encryption of Plain. If Encrypted is bound to a list holding 2 characters and an unbound tail, these two character are used for the salt of the encryption. Finally, if Encrypted is instantiated to an encrypted password the predicate succeeds iff Encrypted is a valid encryption of Plain.

Plain is either an atom, SWI-Prolog string, list of characters or list of character-codes. It is not advised to use atoms, as this implies the password will be available from the Prolog heap as a defined atom.

8 Memory files

The library(memfile) provides an alternative to temporary files, intended for temporary buffering of data. Memory files in general are faster than temporary files and do not suffer from security riscs or naming conflicts associated with temporary-file management. They do assume proper memory management by the hosting OS and cannot be used to pass data to external processes using a file-name.

There is no limit to the number of memory streams, nor the size of them. However, memory-streams cannot have multiple streams at the same time (i.e. cannot be opened for reading and writing at the same time).

These predicates are first of all intended for building higher-level primitives. See also sformat/3, atom_to_term/3, term_to_atom/2 and the XPCE primitive pce_open/3.

new_memory_file(-Handle)
Create a new memory file and return a unique opaque handle to it.

free_memory_file(+Handle)
Discard the memory file and its contents. If the file is open it is first closed.

open_memory_file(+Handle, +Mode, -Stream)
Open the memory-file. Mode is currently one of read or write. The resulting handling is closed using close/1.

size_memory_file(+Handle, -Bytes)
Return the content-length of the memory-file it Bytes. The file should be closed and contain data.

atom_to_memory_file(+Atom, -Handle)
Turn an atom into a read-only memory-file containing the (shared) characters of the atom. Opening this memory-file in mode write yields a permission error.

memory_file_to_atom(+Handle, -Atom)
Return the content of the memory-file in Atom.

memory_file_to_codes(+Handle, -Codes)
Return the content of the memory-file as a list of character-codes in Codes.

9 Time and alarm library

The library(time) provides timing and alarm functions.

alarm(+Time, :Callable, -Id, +Option)
Schedule Callable to be called Time seconds from now. Time is a number (integer or float). Callable is called on the next pass through a call- or redo-port of the Prolog engine, or a call to the PL_handle_signals() routine from SWI-Prolog. Id is unified with a reference to the timer.

The resolution of the alarm depends on the underlying implementation. On Unix systems it is based on setitimer(), on Windows on timeSetEvent() using a resolution specified at 50 milliseconds. Long-running foreign predicates that do not call PL_handle_signals() may further delay the alarm.

Options is a list of Name(Value) terms. Defined options are:

remove(Bool)
If true (default false), the timer is removed automatically. Otherwise it must be delated explicitly using remove_alarm/1.

alarm(+Time, :Callable, -Id)
Same as alarm(Time, Callable, Id,).

remove_alarm(+Id)
Remove an alarm. If it is not yet fired, it will not be fired any more.

current_alarm(?At, ?:Callable, ?Id, ?Status)
Enumerate the not-yet-removed alarms. Status is one of done if the alarm has been called, next if it is the next to be fired and scheduled otherwise.

call_with_time_limit(+Time, :Goal)
Call Goal as once/1. If Goal doesn't complete within Time seconds, exit using the exception time_limit_exceeded. See catch/3.

Please note that this predicate uses alarm/4 and therefore is not capable to break out of long running goals such as sleep/1, blocking I/O or other long-running (foreign) predicates. Blocking I/O can be handled using the timeout option of read_term/3.

10 Limiting process resources

The library(rlimit) library provides an interface to the POSIX getrlimit()/setrlimit() API that control the maximum resource-usage of a process or group of processes. This call is especially useful for server such as CGI scripts and inetd-controlled servers to avoid an uncontrolled script claiming too much resources.

rlimit(+Resource, -Old, +New)
Query and/or set the limit for Resource. Time-values are in seconds and size-values are counted in bytes. The following values are supported by this library. Please note that not all resources may be available and accessible on all platforms. This predicate can throw a variety of exceptions. In portable code this should be guarded with catch/3. The defined resources are:
cpu CPU time in seconds
fsize Maximum filesize
data max data size
stack max stack size
core max core file size
rss max resident set size
nproc max number of processes
nofile max number of open files
memlock max locked-in-memory address

When the process hits a limit POSIX systems normally send the process a signal that terminates it. These signals may be catched using SWI-Prolog's on_signal/3 primitive. The code below illustrates this behaviour. Please note that asynchronous signal handling is dangerous, especially when using threads. 100% fail-safe operation cannot be guaranteed, but this procedure will inform the user properly `most of the time'.


rlimit_demo :-
        rlimit(cpu, _, 2),
        on_signal(xcpu, _, cpu_exceeded),
        ( repeat, fail ).

cpu_exceeded(_Sig) :-
        format(user_error, 'CPU time exceeded~n', []),
        halt(1).

11 Installation

11.1 Unix systems

Installation on Unix system uses the commonly found configure, make and make install sequence. SWI-Prolog should be installed before building this package. If SWI-Prolog is not installed as pl, the environment variable PL must be set to the name of the SWI-Prolog executable. Installation is now accomplished using:


% ./configure
% make
% make install

This installs the foreign libraries in $PLBASE/lib/$PLARCH and the Prolog library files in $PLBASE/library, where $PLBASE refers to the SWI-Prolog `home-directory'.

Footnotes

note-1
kill/2 should support interrupt-names as well
note-2
More subtle handling of I/O, especially for debugging is required: communicate with the syslog deamon and optionally start a debugging dialog on a newly created (X-)terminal should be considered.
note-3
This is hard to achieve at the moment as none of the Prolog read-commands provide for a timeout.
note-4
It is unclear to me whether a MIME note can contain a mixture of content and parts, but I believe the answer is `no'.

Index

A
add_stream_to_pool/2
4.3
alarm/3
alarm/4
9
at_end_of_stream/1
4 4
atom_to_memory_file/2
atom_to_term/3
8
C
call_with_time_limit/2
catch/3
9 10
cgi_get_form/1
close/1
8
close_stream_pool/0
4.3
crypt/2
7
current_alarm/4
D
delete_stream_from_pool/1
detach_IO/0
2
dispatch_stream_pool/1
4.3
dup/2
E
exec/1
F
fork/1
2 2 4 4.1 4.1
fork/2
2
format/2
5.1
free_memory_file/1
G
gethostname/1
H
halt/[0,1]
2
K
kill/2
2
M
memory_file_to_atom/2
memory_file_to_codes/2
mime_parse/2
N
new_memory_file/1
O
once/1
9
on_signal/2
2
on_signal/3
10
open_memory_file/3
open_socket/3
4
P
pce_open/3
8
pipe/2
2 2
R
read/1
4
read_term/3
9
remove_alarm/1
9
rlimit/3
S
set_time_file/3
sformat/3
8
size_memory_file/2
sleep/1
9
socket
4
stream_pool_main_loop/0
T
tcp_accept/3
4 4 4 4 4.1
tcp_bind/2
4 4
tcp_close_socket/1
4 4 4
tcp_connect/2
tcp_fcntl/3
4.2
tcp_host_to_address/2
tcp_listen/2
4
tcp_open_socket/3
4
tcp_setopt/2
tcp_socket/1
term_to_atom/2
8
W
wait/2
2
wait_for_input/3
4.1 4.2 4.3 4.3