932 lines
31 KiB
XML
932 lines
31 KiB
XML
<?xml version="1.0" encoding="ISO-8859-1"?>
|
|
<package xmlns="http://www.lovullo.com/rater"
|
|
xmlns:c="http://www.lovullo.com/calc"
|
|
xmlns:t="http://www.lovullo.com/rater/apply-template"
|
|
core="true"
|
|
desc="Functions for performing table lookups">
|
|
|
|
<import package="../base" />
|
|
<import package="../numeric/common" />
|
|
<import package="list" />
|
|
|
|
<!-- since templates are inlined, we need to make these symbols available to
|
|
avoid terrible confusion -->
|
|
<import package="common" export="true" />
|
|
<import package="matrix" export="true" />
|
|
|
|
|
|
<const name="TABLE_WHEN_MASK_VALUE" type="float" value="-0.0001" desc="Used to mask when conditions" />
|
|
<const name="MFILTER_BISECT_GAP_MAX" type="integer" value="10" desc="Quit bisect if size is less than or equal to this value" />
|
|
|
|
|
|
<!--
|
|
Create a constant table
|
|
|
|
This definition must appear within a `constants' block.
|
|
|
|
Permitted children:
|
|
- _table_column_+ - Column definitions
|
|
- _table_rows_ - Begin table data definition
|
|
-->
|
|
<template name="_create-table_"
|
|
desc="Create an arbitrary table for querying">
|
|
<param name="@name@" desc="Table name" />
|
|
<param name="@values@" desc="Table definition" />
|
|
|
|
<param name="@desc@" desc="Table description">
|
|
<text></text>
|
|
</param>
|
|
|
|
<param name="@__tid@"
|
|
desc="Internal table identifier">
|
|
<param-value name="@name@" upper="true" snake="true" />
|
|
</param>
|
|
|
|
|
|
<param-copy name="@values@">
|
|
<param-meta name="create-table-id" value="@__tid@" />
|
|
<param-meta name="create-table-name" value="@name@" />
|
|
<param-meta name="create-table-desc" value="@desc@" />
|
|
</param-copy>
|
|
</template>
|
|
|
|
|
|
<!--
|
|
Declare a table column name
|
|
|
|
A column definition assigns a field name to a column index. If the
|
|
column always contains ordered (sequenced) data, then @seq@ should
|
|
be set to "true"; this allows a more efficient query strategy to
|
|
be used.
|
|
|
|
If a table is especially large, the first column should be treated
|
|
as the index and always be sequenced.
|
|
-->
|
|
<template name="_table-column_"
|
|
desc="Declare name for table column">
|
|
<param name="@name@" desc="Column name" />
|
|
<param name="@index@" desc="Column index (0-indexed)" />
|
|
|
|
<param name="@desc@" desc="Column description">
|
|
<param-value name="@name@" />
|
|
</param>
|
|
|
|
<!-- use carefully; leave alone unless data is definately sorted,
|
|
or query results may be incorrect -->
|
|
<param name="@seq@" desc="Column is sorted (sequential)">
|
|
<text></text>
|
|
</param>
|
|
|
|
<param name="@__tid@"
|
|
desc="Internal table identifier">
|
|
<param-inherit meta="create-table-id" />
|
|
</param>
|
|
|
|
<!-- FIXME: this doesn't contain @__tid@ because of a bug in the
|
|
preprocessor; this is fixed in the new DSL -->
|
|
<param name="@__constname@"
|
|
desc="Name of internal constant used for column lookup">
|
|
<text>RATE_TABLE_</text>
|
|
<param-value name="@name@" upper="true" snake="true" />
|
|
</param>
|
|
|
|
|
|
<!-- column index identifier -->
|
|
<const name="{@__tid@}_{@__constname@}"
|
|
value="@index@" type="integer"
|
|
desc="@desc@" />
|
|
|
|
<!-- column sequential flag to permit query optimizations -->
|
|
<if name="@seq@" eq="true">
|
|
<const name="{@__tid@}_{@__constname@}_IS_SEQ"
|
|
value="1" type="integer"
|
|
desc="{@name@} is sequenced" />
|
|
</if>
|
|
<unless name="@seq@" eq="true">
|
|
<const name="{@__tid@}_{@__constname@}_IS_SEQ"
|
|
value="0" type="integer"
|
|
desc="{@name@} is unordered" />
|
|
</unless>
|
|
</template>
|
|
|
|
|
|
<!--
|
|
Begin table data definition (rows)
|
|
|
|
Each _table-row_ child inserts a constant row into the table. Note
|
|
that all data must be constant.
|
|
|
|
Permitted children:
|
|
- _table-row_* - Table row definitions
|
|
-->
|
|
<template name="_table-rows_"
|
|
desc="Begin table data definition">
|
|
<param name="@values@" desc="Row definitions" />
|
|
|
|
<param name="@__tid@"
|
|
desc="Table identifier">
|
|
<param-inherit meta="create-table-id" />
|
|
</param>
|
|
<param name="@__tname@"
|
|
desc="Table name as provided by caller">
|
|
<param-inherit meta="create-table-name" />
|
|
</param>
|
|
<param name="@__desc@"
|
|
desc="Table description">
|
|
<param-inherit meta="create-table-desc" />
|
|
</param>
|
|
|
|
|
|
<const name="{@__tid@}_RATE_TABLE"
|
|
type="float"
|
|
desc="{@__tname@} table; {@__desc@}">
|
|
<param-copy name="@values@" />
|
|
</const>
|
|
</template>
|
|
|
|
|
|
<!--
|
|
Define a constant table row
|
|
|
|
Rows will be inserted into the table in the order in which they
|
|
appear. Note that, if the column is marked as sequenced, it is
|
|
important that the order is itself sequenced.
|
|
|
|
All values must be constant.
|
|
|
|
Permitted children:
|
|
_table-value_+ - Row column value
|
|
-->
|
|
<template name="_table-row_"
|
|
desc="Define a constant table row (ordered)">
|
|
<param name="@values@" desc="Column values" />
|
|
|
|
|
|
<set desc="Row">
|
|
<param-copy name="@values@" />
|
|
</set>
|
|
</template>
|
|
|
|
|
|
<!--
|
|
Set a column value for the parent row
|
|
|
|
Column value order should match the defined column order. All
|
|
values must be constants.
|
|
-->
|
|
<template name="_table-value_"
|
|
desc="Table column value for row (ordered)">
|
|
<param name="@const@" desc="Constant column value" />
|
|
|
|
|
|
<item value="@const@" desc="Column value" />
|
|
</template>
|
|
|
|
|
|
<template name="_query-first-field_" desc="Return the requested field from the first row returned by a query">
|
|
<param name="@table@" desc="Table (matrix) to query" />
|
|
<param name="@values@" desc="Query parameters" />
|
|
|
|
<!-- use either field or name[+index] -->
|
|
<param name="@field@" desc="Column to select (variable/constant); do not use with @name@" />
|
|
<param name="@name@" desc="Name of field to query (variable/constant); overrides @field@" />
|
|
<param name="@index@" desc="Optional index for field name lookup (using @name@)" />
|
|
|
|
|
|
<c:car label="First row of query result">
|
|
<t:query-field table="@table@" field="@field@" name="@name@" index="@index@">
|
|
<param-copy name="@values@" />
|
|
</t:query-field>
|
|
</c:car>
|
|
</template>
|
|
|
|
|
|
<!--
|
|
Query-style syntax for matrix searches
|
|
|
|
Overhead of this algorithm is minimal. Let us consider that a direct access
|
|
to an element in a vector is O(1). Therefore, given a set of p predicates
|
|
in a direct matrix access, the time would be O(p).
|
|
|
|
In the case of this query method, we perform first a bisect and then a
|
|
linear search backward to the first element matching the given predicate.
|
|
Therefore, per field, the best case average is O(lg n), where n is the
|
|
number of rows; this assumes that no backward linear search is required.
|
|
Should such a search be required, the worst case per field is O(lg n + m),
|
|
where m is the length of the largest subset.
|
|
|
|
Once again considering p predicates, each predicate P_k where k > 0
|
|
(0-indexed) will operate on a subset of the first predicate. As already
|
|
discussed, the worst-case scenerio for the length of the subset is m. Since
|
|
further queries on this subset are rarely likely to use the bisect
|
|
algorithm, we're looking at O(m) time per field. Therefore, the total query
|
|
time for the best-case scenerio is still O(lg n) if m is sufficiently small
|
|
and O(lg n + pm) if m is sufficiently large.
|
|
|
|
An important case to note is when m approaches (or is equal to) n; in such
|
|
a case, the algorithm degrades to a worst-case linear search of O(pn) and
|
|
best case of O(n) if early predicates are sufficiently effective at
|
|
reducing the subsets to further predicates: That is, the bisect algorithm
|
|
either cannot be used or is ineffective. For large m, this may cause a
|
|
stack overflow due to the recursive nature of the algorithm. Is is
|
|
therefore important to order the first column of the table such that it is
|
|
both sorted and produces small m. Additionally, it is ideal for the first
|
|
predicate to query the first field to quickly reduce the size of the set
|
|
for the next predicate.
|
|
-->
|
|
<template name="_query-field_" desc="Return the requested field from rows returned by a query">
|
|
<param name="@table@" desc="Table (matrix) to query" />
|
|
<param name="@values@" desc="Query parameters" />
|
|
|
|
<!-- use one or the other -->
|
|
<param name="@field@" desc="Column to select (variable/constant); do not use with @name@" />
|
|
|
|
<param name="@index@" desc="Optional index for field name lookup (using @name@)" />
|
|
<!-- by default, if @field@ is provided instead of @name@, the field
|
|
constant will be generated (same concept as the 'when' template) -->
|
|
<param name="@name@" desc="Name of field to query (variable/constant); overrides @field@">
|
|
<!-- convert @table@ to uppercase and snake case (replace - with _) -->
|
|
<param-value name="@table@" upper="true" snake="true" />
|
|
<text>_RATE_TABLE_</text>
|
|
<param-value name="@field@" upper="true" snake="true" />
|
|
</param>
|
|
|
|
|
|
<c:apply name="mcol" label="Query result">
|
|
<!-- the matrix (vector of rows) returned by the query -->
|
|
<c:arg name="matrix">
|
|
<t:query-row table="@table@">
|
|
<param-copy name="@values@" />
|
|
</t:query-row>
|
|
</c:arg>
|
|
|
|
<!-- the field (column) to retrieve; 0-based index -->
|
|
<c:arg name="col">
|
|
<!-- no index lookup needed -->
|
|
<unless name="@index@">
|
|
<c:value-of name="@name@" />
|
|
</unless>
|
|
|
|
<!-- index lookup required -->
|
|
<if name="@index@">
|
|
<c:value-of name="@name@" index="@index@" />
|
|
</if>
|
|
</c:arg>
|
|
</c:apply>
|
|
</template>
|
|
|
|
|
|
<template name="_query-row_" desc="Query a table (matrix) for a row (vector) of values">
|
|
<param name="@table@" desc="Table (matrix)" />
|
|
<param name="@values@" desc="Query parameters" />
|
|
|
|
<!-- this defaults to a table name constant as generated from the csv2xml
|
|
script; either this or @table@ should be used -->
|
|
<param name="@matrix@" desc="Matrix to look up from">
|
|
<!-- convert @table@ to uppercase and snake case (replace - with _) -->
|
|
<param-value name="@table@" upper="true" snake="true" />
|
|
<text>_RATE_TABLE</text>
|
|
</param>
|
|
|
|
<c:apply name="_mquery">
|
|
<c:arg name="matrix">
|
|
<c:value-of name="@matrix@" />
|
|
</c:arg>
|
|
|
|
<c:arg name="criteria">
|
|
<c:set>
|
|
<param-copy name="@values@">
|
|
<param-meta name="table_basename" value="@matrix@" />
|
|
</param-copy>
|
|
</c:set>
|
|
</c:arg>
|
|
|
|
<c:arg name="i">
|
|
<!-- begin with the last predicate (due to the way we'll recurse, it
|
|
will be applied *last* -->
|
|
<t:dec>
|
|
<c:length-of>
|
|
<c:set>
|
|
<param-copy name="@values@">
|
|
<param-meta name="table_basename" value="@matrix@" />
|
|
</param-copy>
|
|
</c:set>
|
|
</c:length-of>
|
|
</t:dec>
|
|
</c:arg>
|
|
</c:apply>
|
|
</template>
|
|
|
|
|
|
<template name="_when_" desc="Create field predicate for query definition">
|
|
<param name="@id@" desc="Field index" />
|
|
<param name="@values@" desc="Field value (provide only one node)" />
|
|
<param name="@sequential@" desc="Is data sequential?" />
|
|
|
|
<!-- @name@ may be provided directly, or @field@ may be used when the
|
|
basename is available (set by a query template), giving the illusion of
|
|
querying the table columns by name directly (magic!); pure syntatic
|
|
sugary goodness -->
|
|
<param name="@field@" desc="Field name (to be used with base)" />
|
|
<param name="@name@" desc="Field name (as a variable/constant)">
|
|
<param-inherit meta="table_basename" />
|
|
<text>_</text>
|
|
<param-value name="@field@" upper="true" snake="true" />
|
|
</param>
|
|
|
|
<param name="@seqvar@" desc="Var/constant containing whether field is sequential">
|
|
<param-inherit meta="table_basename" />
|
|
<text>_</text>
|
|
<param-value name="@field@" upper="true" snake="true" />
|
|
<text>_IS_SEQ</text>
|
|
</param>
|
|
|
|
<c:set label="Conditional for {@field@}">
|
|
<!-- the first element will represent the column (field) index -->
|
|
<unless name="@name@">
|
|
<c:const value="@id@" type="integer" desc="Field index" />
|
|
</unless>
|
|
<if name="@name@">
|
|
<c:value-of name="@name@" />
|
|
</if>
|
|
|
|
<!-- the second element will represent the expected value(s) -->
|
|
<c:set>
|
|
<param-copy name="@values@" />
|
|
</c:set>
|
|
|
|
<!-- the final element will represent whether or not this field is sequential -->
|
|
<if name="@sequential@">
|
|
<c:const value="@sequential@" type="boolean" desc="Whether data is sequential" />
|
|
</if>
|
|
<unless name="@sequential@">
|
|
<!-- if a field name was given, we can get the sequential information
|
|
that was already generated for us -->
|
|
<if name="@field@">
|
|
<c:value-of name="@seqvar@" />
|
|
</if>
|
|
<!-- otherwise, default to non-sequential -->
|
|
<unless name="@field@">
|
|
<c:value-of name="FALSE" />
|
|
</unless>
|
|
</unless>
|
|
</c:set>
|
|
</template>
|
|
|
|
|
|
|
|
<!--
|
|
These functions make the magic happen
|
|
|
|
They are hideous. Look away.
|
|
-->
|
|
|
|
|
|
<!-- this function is intended to be called by the _query_ template, not directly -->
|
|
<function name="_mquery" desc="Query for vectors using a set of column criteria">
|
|
<param name="matrix" type="float" set="matrix" desc="Matrix to query" />
|
|
<param name="criteria" type="float" set="matrix" desc="Query criteria" />
|
|
|
|
<param name="i" type="integer" desc="Current criteria index" />
|
|
|
|
<c:cases>
|
|
<c:case>
|
|
<c:when name="i">
|
|
<c:eq>
|
|
<!-- it's important that we allow index 0, since that is a valid
|
|
predicate -->
|
|
<c:const value="-1" type="integer" desc="We're done." />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<!-- we're done; stick with the result -->
|
|
<c:value-of name="matrix" />
|
|
</c:case>
|
|
|
|
<c:otherwise>
|
|
<c:apply name="mfilter">
|
|
<!-- matrix to search -->
|
|
<c:arg name="matrix">
|
|
<!-- >> recursion happens here << -->
|
|
<c:apply name="_mquery">
|
|
<c:arg name="matrix">
|
|
<c:value-of name="matrix" />
|
|
</c:arg>
|
|
|
|
<c:arg name="criteria">
|
|
<c:value-of name="criteria" />
|
|
</c:arg>
|
|
|
|
<c:arg name="i">
|
|
<t:dec>
|
|
<c:value-of name="i" />
|
|
</t:dec>
|
|
</c:arg>
|
|
</c:apply>
|
|
</c:arg>
|
|
|
|
<!-- field (column) -->
|
|
<c:arg name="col">
|
|
<c:value-of name="criteria">
|
|
<c:index>
|
|
<c:value-of name="i" />
|
|
</c:index>
|
|
<c:index>
|
|
<c:const value="0" type="integer" desc="Field id" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:arg>
|
|
|
|
<!-- value(s) to search for -->
|
|
<c:arg name="vals">
|
|
<c:value-of name="criteria">
|
|
<c:index>
|
|
<c:value-of name="i" />
|
|
</c:index>
|
|
<c:index>
|
|
<c:const value="1" type="integer" desc="Field value" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:arg>
|
|
|
|
<!-- if it's sequential, we can cut down on the search substantially -->
|
|
<c:arg name="seq">
|
|
<c:value-of name="criteria">
|
|
<c:index>
|
|
<c:value-of name="i" />
|
|
</c:index>
|
|
<c:index>
|
|
<c:const value="2" type="integer" desc="Sequential flag" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:arg>
|
|
</c:apply>
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</function>
|
|
|
|
|
|
<!--
|
|
Perform a lg(n) bisect on a data set.
|
|
|
|
This is intended to limit recursion on very large data sets (and
|
|
consequently will increase performance). This will bisect up until a certain
|
|
point (the gap), unless it finds the value in question. After finding the
|
|
value, it will perform a linear backward search to find the first occurrence
|
|
of the value. If the value is not found, it will halt at the gap and return
|
|
the first index of the gap, which we will consider its "best guess", at
|
|
which point a linear search can be performed by the caller to determine if
|
|
the value does in fact exist at all.
|
|
|
|
(The reason this operates so oddly is because of its caller. We could rid
|
|
the gap and make this a general-purpose function if need be. Technically,
|
|
the gap is useless and saves lg(g) steps, which may be very small.)
|
|
-->
|
|
<function name="bisect" desc="Bisect a matrix toward the requested column value">
|
|
<param name="matrix" type="float" set="matrix" desc="Matrix to bisect" />
|
|
<param name="col" type="integer" desc="Column index to filter on" />
|
|
<param name="val" type="float" desc="Column value to filter on" />
|
|
<param name="start" type="integer" desc="Start index" />
|
|
<param name="end" type="integer" desc="Start end" />
|
|
|
|
<c:let>
|
|
<c:values>
|
|
<!-- the gap represents the number of indexes between the current start
|
|
and end indexes -->
|
|
<c:value name="gap" type="integer" desc="Gap between current start and end">
|
|
<c:sum>
|
|
<c:value-of name="end" />
|
|
|
|
<t:negate>
|
|
<c:value-of name="start" />
|
|
</t:negate>
|
|
</c:sum>
|
|
</c:value>
|
|
</c:values>
|
|
|
|
<!--
|
|
At this point, we need to determine if we should continue the bisect or
|
|
halt. The purpose of the gap is based on the understanding that (with
|
|
our use cases) we will arrive, almost always, at one of two scenarios:
|
|
we found the value, but it's part of a larger set of the same values, or
|
|
the value we are looking for may not even exist at all.
|
|
|
|
The gap just limits recursion (but just barely) at smaller levels, since
|
|
bisect will take lg(n) steps). Increase the gap limit to decrease the
|
|
number of steps, or decrease it to 1 if you want a complete bisection.
|
|
-->
|
|
<c:cases>
|
|
<!-- give up if we've reached our gap limit -->
|
|
<c:case>
|
|
<c:when name="gap">
|
|
<c:lte>
|
|
<c:value-of name="MFILTER_BISECT_GAP_MAX" />
|
|
</c:lte>
|
|
</c:when>
|
|
|
|
<!-- we tried our best; return our current position -->
|
|
<c:value-of name="start" />
|
|
</c:case>
|
|
|
|
<!-- we have not yet reached our gap limit; keep going -->
|
|
<c:otherwise>
|
|
<c:let>
|
|
<c:values>
|
|
<c:value name="mid_index" type="integer" desc="Middle index">
|
|
<!-- to determine the new mid index, add half of the gap to the
|
|
current index -->
|
|
<c:sum>
|
|
<c:value-of name="start" />
|
|
<c:ceil>
|
|
<c:quotient>
|
|
<c:value-of name="gap" />
|
|
<c:const value="2" type="integer" desc="Bisect" />
|
|
</c:quotient>
|
|
</c:ceil>
|
|
</c:sum>
|
|
</c:value>
|
|
</c:values>
|
|
|
|
<c:let>
|
|
<c:values>
|
|
<c:value name="mid" type="float" desc="Middle value">
|
|
<c:value-of name="matrix">
|
|
<!-- row -->
|
|
<c:index>
|
|
<c:value-of name="mid_index" />
|
|
</c:index>
|
|
|
|
<!-- column -->
|
|
<c:index>
|
|
<c:value-of name="col" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:value>
|
|
</c:values>
|
|
|
|
<c:cases>
|
|
<!-- if the middle value is lower than our value, then take the upper half -->
|
|
<c:case>
|
|
<c:when name="mid">
|
|
<c:lt>
|
|
<c:value-of name="val" />
|
|
</c:lt>
|
|
</c:when>
|
|
|
|
<c:recurse start="mid_index" />
|
|
</c:case>
|
|
|
|
<!-- similarily, take the lower half if we over-shot -->
|
|
<c:case>
|
|
<c:when name="mid">
|
|
<c:gt>
|
|
<c:value-of name="val" />
|
|
</c:gt>
|
|
</c:when>
|
|
|
|
<c:recurse end="mid_index" />
|
|
</c:case>
|
|
|
|
<!-- if we have an exact match, that doesn't necessarily mean that we
|
|
have every value; we may have intersected a set of them -->
|
|
<c:otherwise>
|
|
<!-- this will return an exact index: the first index
|
|
containing the element we've been looking for (it is a
|
|
linear backwards search) -->
|
|
<c:apply name="foremost" matrix="matrix" col="col" i="mid_index" />
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</c:let>
|
|
</c:let>
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</c:let>
|
|
</function>
|
|
|
|
|
|
<function name="foremost" desc="Search backwards for the first occurrance in a sorted list">
|
|
<param name="matrix" type="float" set="matrix" desc="Matrix to bisect" />
|
|
<param name="col" type="integer" desc="Column index to search on" />
|
|
<param name="i" type="integer" desc="Current index" />
|
|
|
|
<c:let>
|
|
<c:values>
|
|
<!-- we calculate this rather than accept it via an argument so that
|
|
this function may be called directly in a more convenient manner
|
|
-->
|
|
<c:value name="val" type="float" desc="Current value">
|
|
<c:value-of name="matrix">
|
|
<!-- row -->
|
|
<c:index>
|
|
<c:value-of name="i" />
|
|
</c:index>
|
|
|
|
<!-- column -->
|
|
<c:index>
|
|
<c:value-of name="col" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:value>
|
|
|
|
<c:value name="prev" type="float" desc="Previous value">
|
|
<c:value-of name="matrix">
|
|
<!-- row -->
|
|
<c:index>
|
|
<t:dec>
|
|
<c:value-of name="i" />
|
|
</t:dec>
|
|
</c:index>
|
|
|
|
<!-- column -->
|
|
<c:index>
|
|
<c:value-of name="col" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:value>
|
|
</c:values>
|
|
|
|
<c:cases>
|
|
<!-- if we have no more indexes to check, then we're done -->
|
|
<c:case>
|
|
<c:when name="i">
|
|
<c:eq>
|
|
<c:const value="0" type="integer" desc="Did we check the final (first) index?" />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<!-- well, then, we're done -->
|
|
<c:value-of name="i" />
|
|
</c:case>
|
|
|
|
<!-- if the previous column value is the same value, then continue checking -->
|
|
<c:case>
|
|
<c:when name="prev">
|
|
<c:eq>
|
|
<c:value-of name="val" />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<c:recurse>
|
|
<c:arg name="i">
|
|
<t:dec>
|
|
<c:value-of name="i" />
|
|
</t:dec>
|
|
</c:arg>
|
|
</c:recurse>
|
|
</c:case>
|
|
|
|
<!-- otherwise, we've found the foremost index -->
|
|
<c:otherwise>
|
|
<c:value-of name="i" />
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</c:let>
|
|
</function>
|
|
|
|
|
|
<template name="_mask-unless_" desc="Mask a value unless the condition is truthful">
|
|
<param name="@values@" desc="Body" />
|
|
<param name="@name@" desc="Scalar to check" />
|
|
<param name="@index@" desc="Optional index" />
|
|
<param name="@desc@" desc="Optional description" />
|
|
|
|
<c:cases>
|
|
<!-- if masked -->
|
|
<c:case>
|
|
<!-- no index provided -->
|
|
<unless name="@index@">
|
|
<c:when name="@name@">
|
|
<c:eq>
|
|
<c:value-of name="FALSE" />
|
|
</c:eq>
|
|
</c:when>
|
|
</unless>
|
|
|
|
<!-- index provided -->
|
|
<if name="@index@">
|
|
<c:when name="@name@" index="@index@">
|
|
<c:eq>
|
|
<c:value-of name="FALSE" />
|
|
</c:eq>
|
|
</c:when>
|
|
</if>
|
|
|
|
<!-- TODO: configurable mask via meta and/or param -->
|
|
<c:value-of name="TABLE_WHEN_MASK_VALUE" />
|
|
</c:case>
|
|
|
|
<!-- if not masked -->
|
|
<c:otherwise>
|
|
<param-copy name="@values@" />
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</template>
|
|
|
|
|
|
<function name="mfilter" desc="Filter matrix rows by column value">
|
|
<param name="matrix" type="float" set="matrix" desc="Matrix to filter" />
|
|
<param name="col" type="integer" desc="Column index to filter on" />
|
|
<param name="vals" type="float" desc="Column value to filter on" />
|
|
<param name="seq" type="boolean" desc="Is data sequential?" />
|
|
|
|
<!-- merge the result of each condition in vals into a single set, which
|
|
has the effect of supporting multiple conditions on a single column of
|
|
data (or just one, if they don't want to feel sweet). By performing
|
|
the lookups separately for each, we preserve the bisect-ability of the
|
|
condition. -->
|
|
<t:merge-until-empty set="vals" car="val" glance="TABLE_WHEN_MASK_VALUE">
|
|
<c:apply name="range" matrix="matrix" col="col" val="val" seq="seq">
|
|
<c:arg name="start">
|
|
<c:cases>
|
|
<!-- if we know that the data is sequential, then we may not need to
|
|
perform a linear search (if the dataset is large enough and the
|
|
column value is relatively distinct) -->
|
|
<c:case>
|
|
<c:when name="seq">
|
|
<c:eq>
|
|
<c:value-of name="TRUE" />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<c:apply name="bisect" matrix="matrix" col="col" val="val">
|
|
<c:arg name="start">
|
|
<c:const value="0" type="integer" desc="Start bisect at beginning" />
|
|
</c:arg>
|
|
|
|
<c:arg name="end">
|
|
<!-- bisect the length of the matrix -->
|
|
<t:dec>
|
|
<c:length-of>
|
|
<c:value-of name="matrix" />
|
|
</c:length-of>
|
|
</t:dec>
|
|
</c:arg>
|
|
</c:apply>
|
|
</c:case>
|
|
|
|
<!-- we have no good guess; linear search :x -->
|
|
<c:otherwise>
|
|
<c:const value="0" type="integer" desc="Start at the first element" />
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</c:arg>
|
|
|
|
<c:arg name="end">
|
|
<t:dec>
|
|
<c:length-of>
|
|
<c:value-of name="matrix" />
|
|
</c:length-of>
|
|
</t:dec>
|
|
</c:arg>
|
|
</c:apply>
|
|
</t:merge-until-empty>
|
|
</function>
|
|
|
|
|
|
<function name="range" desc="Filter matrix rows by column value within a certain range of indexes (inclusive)">
|
|
<param name="matrix" type="float" set="matrix" desc="Matrix to filter" />
|
|
<param name="col" type="integer" desc="Column index to filter on" />
|
|
<param name="val" type="float" desc="Column value to filter on" />
|
|
<param name="start" type="integer" desc="Starting index (inclusive)" />
|
|
<param name="end" type="integer" desc="Ending index (inclusive)" />
|
|
<param name="seq" type="boolean" desc="Is data sequential?" />
|
|
|
|
<c:let>
|
|
<c:values>
|
|
<c:value name="curval" type="float" desc="Current value">
|
|
<c:value-of name="matrix">
|
|
<c:index>
|
|
<c:value-of name="start" />
|
|
</c:index>
|
|
|
|
<c:index>
|
|
<c:value-of name="col" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:value>
|
|
</c:values>
|
|
|
|
<!-- nested let needed so that the curval is available to over
|
|
in the body below -->
|
|
<c:let>
|
|
<c:values>
|
|
<!-- determine if the value we're looking for is over the current value
|
|
in a sorted list (meaning that we will not find it) -->
|
|
<c:value name="over" type="boolean" desc="Did we pass the potential value in a sorted list?">
|
|
<c:value-of name="TRUE">
|
|
<c:when name="seq">
|
|
<c:eq>
|
|
<c:value-of name="TRUE" />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<c:when name="curval">
|
|
<c:gt>
|
|
<c:value-of name="val" />
|
|
</c:gt>
|
|
</c:when>
|
|
</c:value-of>
|
|
</c:value>
|
|
</c:values>
|
|
|
|
<c:cases>
|
|
<!-- if we're done filtering, then return an empty set -->
|
|
<c:case>
|
|
<c:when name="start">
|
|
<c:gt>
|
|
<c:value-of name="end" />
|
|
</c:gt>
|
|
</c:when>
|
|
|
|
<!-- empty set -->
|
|
<c:set />
|
|
</c:case>
|
|
|
|
<!-- if the data is sequential and the next element is over the
|
|
requested value, then we're done -->
|
|
<c:case>
|
|
<c:when name="over">
|
|
<c:eq>
|
|
<c:value-of name="TRUE" />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<!-- empty set -->
|
|
<c:set />
|
|
</c:case>
|
|
|
|
|
|
<c:otherwise>
|
|
<c:apply name="_mfilter" matrix="matrix" col="col" val="val" start="start" end="end" seq="seq">
|
|
<c:arg name="cur">
|
|
<c:value-of name="matrix">
|
|
<!-- current row -->
|
|
<c:index>
|
|
<c:value-of name="start" />
|
|
</c:index>
|
|
|
|
<!-- requested column -->
|
|
<c:index>
|
|
<c:value-of name="col" />
|
|
</c:index>
|
|
</c:value-of>
|
|
</c:arg>
|
|
</c:apply>
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</c:let>
|
|
</c:let>
|
|
</function>
|
|
|
|
|
|
<function name="_mfilter" desc="mfilter helper">
|
|
<param name="matrix" type="float" set="matrix" desc="Matrix to filter" />
|
|
<param name="col" type="integer" desc="Column index to filter on" />
|
|
<param name="val" type="float" desc="Column value to filter on" />
|
|
<param name="start" type="integer" desc="Starting index (aka current index)" />
|
|
<param name="end" type="integer" desc="Ending index" />
|
|
<param name="seq" type="integer" desc="Is data sequential?" />
|
|
|
|
<param name="cur" type="float" desc="Current value" />
|
|
|
|
<c:cases>
|
|
<c:case>
|
|
<c:when name="cur">
|
|
<c:eq>
|
|
<c:value-of name="val" />
|
|
</c:eq>
|
|
</c:when>
|
|
|
|
<c:cons>
|
|
<c:value-of name="matrix">
|
|
<c:index>
|
|
<c:value-of name="start" />
|
|
</c:index>
|
|
</c:value-of>
|
|
|
|
<c:apply name="range" matrix="matrix" col="col" val="val" end="end" seq="seq">
|
|
<c:arg name="start">
|
|
<c:sum>
|
|
<c:value-of name="start" />
|
|
<c:const value="1" type="integer" desc="Check next element" />
|
|
</c:sum>
|
|
</c:arg>
|
|
</c:apply>
|
|
</c:cons>
|
|
</c:case>
|
|
|
|
<c:otherwise>
|
|
<c:apply name="range" matrix="matrix" col="col" val="val" end="end" seq="seq">
|
|
<c:arg name="start">
|
|
<c:sum>
|
|
<c:value-of name="start" />
|
|
<c:const value="1" type="integer" desc="Check next element" />
|
|
</c:sum>
|
|
</c:arg>
|
|
</c:apply>
|
|
</c:otherwise>
|
|
</c:cases>
|
|
</function>
|
|
</package>
|
|
|