NAME
Finance::Shares::Code - user programmable control
SYNOPSIS
Three examples of how to specify a test, one showing the minimum required and the other illustrating all the possible fields.
use Finance::Shares::Model;
my @spec = (
...
tests => [
basic => qq( # perl code here ),
iterative => {
before => qq( # perl code here ),
step => qq( # perl code here ),
after => qq( # perl code here ),
},
code => sub { print "hello world\n" },
],
samples => [
...
one => {
tests => ['basic', 'iterative'],
# NOT 'code' !
...
}
],
);
my $fsm = new Finance::Shares::Model( @spec );
$fsm->build();
DESCRIPTION
This module allows model test code to write points or lines on the graphs.
The code must be declared within a code block in a Finance::Shares::Model specification. Each test is a segment of plain text containing perl code. This is compiled by the model engine and run as required.
Some special variables and functions provide links with the rest of the model data. Other lines are referenced using pseudo-variables (e.g. $line_tag
). A function, info()
provides access to the Line and Function objects used, and another, mark()
draws graph lines point by point. Each of these uses create their lines if they don't already exist, so a lines entry is only needed if the default behaviour isn't what you want.
Additional functions include value()
which is used by some Finance::Shares::Functions to return statistics or other calculated results.
Perhaps the function with most possibilities is call()
which allows code to call, at any time, methods/functions provided by other modules.
Code Types
As shown by the SYNOPSIS examples, there are three types of code.
Simple text
The simplest and most common form, this perl fragment is run on every data point in the sample where it is applied. It can be used to produce conditional signal marks on the charts, noting dates and prices for potential stock positions either as printed output or by invoking a callback function.
Hash ref
This may have three keys, before
, step
and after
. Each is a perl fragment in text form. before
is run once, before the sample is considered. step
is run for every point in the sample and is identical to "Simple text". after
is run once after step
has finished.
This form allows one-off invocation of callbacks, an open-print-close sequence for writing signals to file, or the flexibility to write stand-alone tests.
To assist debugging there is an additional field, verbose
, which overrides the global value. Setting this to 2 or more shows the code fragments that are actually evaluated. With '2', the line substitutions are listed by their keys while '3' lists them by their internal names.
Raw code
These are the callbacks invoked by the previous types of perl fragment. The parameters are those used to call the code. See "call" below.
Variables
In addition to perl variables declared with my
or our
, undeclared scalar variables are used to refer to the values of other lines on the chart.
Before the code is compiled a search is made for everything that looks like a scalar or list variable that expands to a line name. The name is then replaced by a reference to the internal data so that it executes correctly.
Pseudo-variables
As in the lines specifications, it is possible to refer to several lines at once. If this is prefixed by '@', all matching values are returned as a list. The same expression prefixed by '$' returns the first value. [Note that it does NOT return the number in the array.]
The special variables are as follows.
- (a)
-
A tag from names or lines;
It is important to avoid all variable names used by Model. At present these are $open, $high, $low, $close and $volume - aliases for the Finance::Shares::data lines.
[Warning Due to the way code fragments are pre-processed, variables should not begin with these names, either.]
- (b)
-
A fully qualified line name such as
sample1/MSFT/date1/my_tag sample1/MSFT/date1/bollinger/high
See "Fully Qualified Line Names" in Finance::Shares::Model for further details.
- (c)
-
A FQLN with wildcards or (some) regular expressions.
As with lines, the wildcard '*' means "any (sample, stock or date) other than this one". Use the regular expression '.*' to match every page including the current one.
[The support for regular expressions is quite limited. Code fragments handle line references by rewriting (like 'C' macro pre-processing). So to use a regular expression as a name, it becomes necessary to search/replace a regexp string, temporarily ignoring any special characters. Only a few of these are escaped at present: '$', '@', '.', '*'. These are just enough to handle lists of lines, the wildcard '*' and regular expression '.*'.]
Otherwise the '$' or '@' value is assumed to be a valid perl variable and is left alone.
Examples
lines => [
avg => {
function => 'moving_average',
},
boll => {
function => 'bollinger_band',
},
],
test => q(
$v1 = $avg; # normal line
$v2 = $boll/high; # one of the 2 produced
@v3 = @*/*/*/close; # i.e. all other close prices
@v4 = @.*/.*//volume; # i.e. all volumes with this date
$v5 = $morrison/MRW.L/default/close; # 'close' is an alias (a 'name')
$v6 = $morrison/MRW.L/default/boll/low; # fully qualified line name
$v7 = $tesco///close; # fqln with current stock & date
@v8 = @data; # all the 'data' lines
$v9 = @data; # the number of 'data' lines
$v0 = $data; # the first 'data' line
),
samples => [
morrison => {
stock => 'MRW.L',
test => 'default',
},
tesco => {
stock => 'TSCO.L',
test => 'default',
},
],
If the variable refers (or defaults) to a line, it yields the value for the line at that date. Clearly this only makes sense while the test is iterating through the data in the step
phase. [If a line is referenced in before
or after
code fragments, they return the Finance::Shares::Line object rather than its data.]
Predefined variables
Each test also has a special variable $self
available. This is a hash ref holding persistent values (similar to an object ref, but not blessed). A couple of variables are declared that change with every iteration. The following values are predefined.
- $date
-
Only useful in the
step
phase, this is set to the current date. - $i
-
The loop counter used in the
step
phase. It indexes the data points. - $self->{by}
-
How the time periods are counted. Will be one of quotes, weekdays, days, weeks, months.
- $self->{code}
-
A hash giving access to all code entries known to the model. Using this, subroutines registered in the code blocks can be called directly.
- $self->{end}
-
The last date normally displayed on the chart.
- $self->{first}
-
The first date which has data available. The
step
phase starts with this. - $self->{last}
-
The last date which has data available. The
step
phase ends with this. - $self->{lines}
-
This array lists all the source lines mentioned in this code fragment.
- $self->{marks}
-
This array lists all the output lines - those referenced within mark() calls.
- $self->{name}
-
The fully qualified line name formed by adding the code tag to the page name. It can be used to provide user input to an anonymous code fragment. See "Portable Code".
- $self->{page}
-
This returns the string identifying the current page.
- $self->{quotes}
-
This is the Finance::Shares::data object used to store the dates, price and volume data for the page. See "Page Names" in Finance::Shares::Model.
It should not normally be needed, but it provides a door into the engine for those who need to access it.
- $self->{start}
-
The first date normally displayed on the chart.
- $self->{stock}
-
The stock code, as given to the model's sample entry for this page.
- $self->{tag}
-
The user defined tag identifying this code entry within a codes block.
Although some effort has been made to make the code appear to be 'natural' perl, it is worth remembering that it is not. For example, any assignment to '$i' will not stick (the model uses it as the iteration counter).
If something won't parse try putting in a space after the pseudo-variable or using concatenation rather than embedding. You can view the code that is actually evaluated - i.e. after the preprocessor substitutions - by setting the verbose
option to 2 or more.
Marking the Chart
A special function is provided which will add a point to a chart line. It is called as
mark( tag, value )
tag
-
This must refer to lines block entry which has
mark
as the function field. [It may be a string literal or a run-time variable. But see "Information on Other Lines".] value
-
A perl expression evaluating to a number or undefined. Often a pseudo-scalar variable refering to a line. See "Variables" above.
Example
lines => [
circle => {
function => 'mark',
},
average => {
function => 'moving_average',
},
],
code => [
show => q(
mark('circle', $average);
),
],
sample => {
test => 'show',
},
In this example the moving average line is marked twice. As well as the normal line, the show
test is invoked and a circle (the default mark) is placed at every position where average
is defined.
Notice that the engine invokes the moving average function for you as it is referred to in the test. There is no need to list it explicitly as a line.
Information on Other Lines
As most code works on data in other lines, it is useful to be able to access information from other Finance::Shares::Function and Finance::Shares::Line objects. This is done using the special function, info.
info( <tag> [, <method> [, <use_fn> ]] )
- <tag>
-
Identifies the line/function in question. It can be a fully qualified line name either as a literal in quotes, a pseudo-variable beginning with '$' or a proper variable.
So the following are legal calls where
tag
is a lines identifier,line
is a suitable line identifier (e.g. 'default') and$fqln
is a variable holding a fully qualified line name (see "Fully Qualified Line Names" in Finance::Shares::Model).info( 'tag' ) info( $tag ) info( 'tag/line' ) info( $tag/line ) info( $fqln )
[WARNING: The variable form is the most problematic. As the line's identity isn't known at compile time, there is no way to ensure it has been built, and it doesn't take part in the internal checks. Also, literal tags can be looked up at compile time, but the variable form must be looked up and every time the code is run. So avoid using variables as identifiers in
step
sections.] - <method>
-
This should be the name of a method from either Finance::Shares::Function or Finance::Shares::Line classes. At present, only methods without arguments can be called, but these are enough for most purposes.
Function methods include:
Method Returns ------------------------------------------------------- id lines tag identifying the specification name FQLN, i.e. sample/stock/date/id chart Finance::Shares::Chart object model Finance::Shares::Model object line_ids list of dependent line tags/fqlns func_lines list of created Finance::Shares::Lines source_lines structured list of FS::Lines used sources flat list of Finance::Shares::Lines value [better to use the value() call] finalize ensure graph axis fit around lines
Line methods include:
Method Returns ------------------------------------------------------- id the line identifier name FQLN, i.e. sample/stock/date/tag/id data array of values for calculations display array of values for display (as data) npoints number of data points function the parent Finance::Shares::Function chart Finance::Shares::Chart object graph hash ref holding graph settings order Z-order display position in graph for_scaling true if line doesn't match Y axis is_mark true if line is generated from code
If
method
is omitted, the object reference itself is returned. - <use_fn>
-
The function can return information on either Lines or Functions. If omitted (or false)
tag
is assumed to be a Finance::Shares::Line object. Setting this to true means that themethod
will be called on the Finance::Shares::Function object that created it.Where
tag
is completely explicit, it refers to a Line and there is no confusion. But iftag
evaluates to a FQLN without a line identifier, this is assumed to be a Function (NOT the 'default' line). If, in this case,use_fn
is not set,method
is called on the FIRST created line. For example,info() call Returns ------------------------------------------------------- info($tag) first FS::Line object info($tag/line) the FS::Line object requested info($tag,,1) the FS::Function object info($tag,'name') FQLN of the first line info($tag/line,'name') calls the Line's 'name' method info($tag,'name',1) the Function's 'name' method
value( <tag> [, <which_value>] )
A poor relation of the other functions, this just calls the value
method of the indicated Finance::Shares::Function. (The parent Function is used if a Line tag is given.) tag
must be a pseudo-variable i.e. a tag name preceded by a '$'.
Some Functions (e.g. sample_mean) just provide a single value, but others (e.g. standard_deviation) produce several. In that case, it is necessary to pass a string indicating which.
my $max = value($max_line);
my $sd = value($sd_line, 'std_dev');
[The line is NOT created when referenced here. This isn't a problem as the Function normally requires a lines entry. But if it is, use the line name as a pseudo-variable or do an info()
or mark()
call on it first to ensure it exists.
[This interface is likely to change so that it catches up with the others.]
Portable Code
The line/function tag may be a variable, provided that variable holds a fully qualified line name. This makes for portable code if used in conjunction with $self-
{name}> which holds the FQLN for the code. For example,
my $func = info($self->{name}, 'function');
my @lines = info($self->{name}, 'func_lines', 1);
These require a lines tag with the same name as the code tag containing it. Thus code can be stored in a file and used in any model, provided the code and line entries use the same tag name.
Example
The file band.code contains a code hash with before
, step
and after
code.
{
before => q(
# Access the 'offset' field given in the lines entry
my $fn = info($self->{name}, 'function');
$self->{offset} = $fn->{offset} || 5;
# Identify the single source line
my @lines = info($self->{name}, 'sources', 1);
$self->{source} = $lines[0];
# Other prior working to keep the step code simple
$self->{high} = $self->{name} . '/high';
$self->{low} = $self->{name} . '/low';
),
step => q(
# The line's value is the i'th in the 'data' array
my $array = $self->{source}->data;
my $value = $array->[$i];
# Only add to the line if the value exists
if (defined $value) {
mark($self->{high}, $value + $self->{offset});
mark($self->{low} , $value - $self->{offset});
}
),
after => q(
# Ensure the rest of the model knows about this data
info($self->{name}, 'finalize', 1);
),
}
A model specification using this might then include:
code => [
my_tag => do('band.code'),
...
],
lines => [
my_tag => {
line => 'open',
offset => 3,
out => ['high', 'low'],
},
],
sample => {
code => ['my_tag'],
line => ['my_tag'],
},
This then produces two lines 3 above and 3 below each date's closing price.
[The field lines=>my_tag=>out
is optional. However, without it a 'default' line will be created at the first info()
call, and it will clutter up the chart's Key.
Likewise, the inclusion of my_tag
in sample=>line
is optional but recommended. Without it, the line is unknown to the system and the dependencies could well become confused.]
Calling Foreign Code
Model code entries may be code refs as well as text. The code is then called using a special function, which will return whatever the code returns.
call( tag, arguments )
scalar = call( tag, arguments )
list = call( tag, arguments )
tag
-
This must be a tag in a tests block identifying a code ref.
arguments
-
This list of arguments (or none) is passed to the subroutine identified by
tag
.
Example
tests => [
tcode => {
my $v = shift;
print "$v->{stock} quotes from $v->{first}\n";
},
doit => {
before => q(
call('tcode', $self);
),
},
],
sample => {
test => 'doit',
},
Remember that $self
has a number of predefined fields, {stock}
and {first}
among them. The before
code is only invoked at the start of the test. It calls the subroutine which prints a message.
This might seem an involved way to go about it, but it is extremely powerful. The subroutine call is made within code which has all the model values (as well as the complete power of perl) available. It may be passed whatever values you choose.
Example
This is a slightly more useful example. It assumes that the MyPortfolio module has exported subroutines defined for simulating buying/selling shares and keeping track of positions you hold in the market. A 'buy' call is made when the stock price seems to rise.
# declare function modules used
use Finance::Shares::moving_average;
# import callback from another module
use MyPortFolio 'enter_long_position';
# Model specification begins
...
lines => [
fast => {
function => 'moving_average',
period => 5,
},
slow => {
function => 'moving_average',
period => 20,
},
],
tests => [
buy => &enter_long_position,
code => q(
call('buy', 'pf1', $self->{date}, $low, $high)
if $fast >= $slow;
);
],
sample => {
test => 'code',
},
When the 5 day moving average crosses above the 20 day one, the callback is invoked. It will be passed a portfolio ID, the date, lowest and highest proces for the day.
<realism> A buy signal would need to be invoked on the basis if at least yesterday's data as todays is usually not yet known! This might be done by using a $self->{buy} flag and having after => 1
set in the appropriate dates entry. </realism>
BUGS
These lines are reserved: $open, $high, $low, $close, $volume. If a test uses a tag name starting with any of these, it will get confused.
Please let me know when you suspect something isn't right. A short script working from a CSV file demonstrating the problem would be very helpful.
AUTHOR
Chris Willmot, chris@willmot.org.uk
LICENCE
Copyright (c) 2002-2003 Christopher P Willmot
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. A copy can be found at http://www.gnu.org/copyleft/gpl.html
SEE ALSO
Finance::Shares::Overview provides an introduction to the suite, and fsmodel is the principal script.
Modules involved in processing the model include Finance::Shares::Model, Finance::Shares::MySQL, Finance::Shares::Chart. Chart and file details may be found in PostScript::File, PostScript::Graph::Paper, PostScript::Graph::Key, PostScript::Graph::Style.
All functions are invoked from their own modules, all with lower-case names such as Finance::Shares::moving_average. The nitty-gritty on how to write each line specification are found there.
The quote data is stored in a Finance::Shares::data object. For information on writing additional line functions see Finance::Shares::Function and Finance::Shares::Line. Also, Finance::Shares::Code covers writing your own tests.