642 lines
21 KiB
C++
642 lines
21 KiB
C++
/*****************************************************************************
|
|
*
|
|
* This file is part of Mapnik (c++ mapping toolkit)
|
|
*
|
|
* Copyright (C) 2014 Artem Pavlenko
|
|
*
|
|
* This library is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2.1 of the License, or (at your option) any later version.
|
|
*
|
|
* This library 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
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with this library; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*
|
|
*****************************************************************************/
|
|
|
|
// NOTE: This is an implementation header file and is only meant to be included
|
|
// from implementation files. It therefore doesn't have an include guard. To
|
|
// create a custom feature_style_processor, include this file and instantiate
|
|
// the template with the desired template arguments.
|
|
|
|
// mapnik
|
|
#include <mapnik/map.hpp>
|
|
#include <mapnik/debug.hpp>
|
|
#include <mapnik/feature.hpp>
|
|
#include <mapnik/feature_style_processor.hpp>
|
|
#include <mapnik/query.hpp>
|
|
#include <mapnik/datasource.hpp>
|
|
#include <mapnik/feature_type_style.hpp>
|
|
#include <mapnik/box2d.hpp>
|
|
#include <mapnik/layer.hpp>
|
|
#include <mapnik/rule.hpp>
|
|
#include <mapnik/rule_cache.hpp>
|
|
#include <mapnik/attribute_collector.hpp>
|
|
#include <mapnik/expression_evaluator.hpp>
|
|
#include <mapnik/scale_denominator.hpp>
|
|
#include <mapnik/projection.hpp>
|
|
#include <mapnik/proj_transform.hpp>
|
|
#include <mapnik/util/featureset_buffer.hpp>
|
|
#include <mapnik/util/variant.hpp>
|
|
#include <mapnik/symbolizer_dispatch.hpp>
|
|
|
|
// stl
|
|
#include <vector>
|
|
#include <stdexcept>
|
|
|
|
namespace mapnik
|
|
{
|
|
|
|
// Store material for layer rendering in a two step process
|
|
struct layer_rendering_material
|
|
{
|
|
layer const& lay_;
|
|
projection const& proj0_;
|
|
projection proj1_;
|
|
box2d<double> layer_ext2_;
|
|
std::vector<feature_type_style const*> active_styles_;
|
|
std::vector<featureset_ptr> featureset_ptr_list_;
|
|
std::vector<rule_cache> rule_caches_;
|
|
|
|
layer_rendering_material(layer const& lay, projection const& dest)
|
|
:
|
|
lay_(lay),
|
|
proj0_(dest),
|
|
proj1_(lay.srs(),true) {}
|
|
};
|
|
|
|
using layer_rendering_material_ptr = std::shared_ptr<layer_rendering_material>;
|
|
|
|
|
|
template <typename Processor>
|
|
feature_style_processor<Processor>::feature_style_processor(Map const& m, double scale_factor)
|
|
: m_(m)
|
|
{
|
|
// https://github.com/mapnik/mapnik/issues/1100
|
|
if (scale_factor <= 0)
|
|
{
|
|
throw std::runtime_error("scale_factor must be greater than 0.0");
|
|
}
|
|
}
|
|
|
|
template <typename Processor>
|
|
void feature_style_processor<Processor>::apply(double scale_denom)
|
|
{
|
|
Processor & p = static_cast<Processor&>(*this);
|
|
p.start_map_processing(m_);
|
|
|
|
projection proj(m_.srs(),true);
|
|
if (scale_denom <= 0.0)
|
|
scale_denom = mapnik::scale_denominator(m_.scale(),proj.is_geographic());
|
|
scale_denom *= p.scale_factor(); // FIXME - we might want to comment this out
|
|
|
|
// Asynchronous query supports:
|
|
// This is a two steps process,
|
|
// first we setup all queries at layer level
|
|
// in a second time, we fetch the results and
|
|
// do the actual rendering
|
|
|
|
std::vector<layer_rendering_material_ptr> mat_list;
|
|
|
|
// Define processing context map used by datasources
|
|
// implementing asynchronous queries
|
|
feature_style_context_map ctx_map;
|
|
|
|
for ( layer const& lyr : m_.layers() )
|
|
{
|
|
if (lyr.visible(scale_denom))
|
|
{
|
|
std::set<std::string> names;
|
|
layer_rendering_material_ptr mat = std::make_shared<layer_rendering_material>(lyr, proj);
|
|
|
|
prepare_layer(*mat,
|
|
ctx_map,
|
|
p,
|
|
m_.scale(),
|
|
scale_denom,
|
|
m_.width(),
|
|
m_.height(),
|
|
m_.get_current_extent(),
|
|
m_.buffer_size(),
|
|
names);
|
|
|
|
// Store active material
|
|
if (!mat->active_styles_.empty())
|
|
{
|
|
mat_list.push_back(mat);
|
|
}
|
|
}
|
|
}
|
|
|
|
for ( layer_rendering_material_ptr mat : mat_list )
|
|
{
|
|
if (!mat->active_styles_.empty())
|
|
{
|
|
render_material(*mat,p);
|
|
}
|
|
}
|
|
|
|
p.end_map_processing(m_);
|
|
}
|
|
|
|
template <typename Processor>
|
|
void feature_style_processor<Processor>::apply(mapnik::layer const& lyr,
|
|
std::set<std::string>& names,
|
|
double scale_denom)
|
|
{
|
|
Processor & p = static_cast<Processor&>(*this);
|
|
p.start_map_processing(m_);
|
|
projection proj(m_.srs(),true);
|
|
if (scale_denom <= 0.0)
|
|
scale_denom = mapnik::scale_denominator(m_.scale(),proj.is_geographic());
|
|
scale_denom *= p.scale_factor();
|
|
|
|
if (lyr.visible(scale_denom))
|
|
{
|
|
apply_to_layer(lyr,
|
|
p,
|
|
proj,
|
|
m_.scale(),
|
|
scale_denom,
|
|
m_.width(),
|
|
m_.height(),
|
|
m_.get_current_extent(),
|
|
m_.buffer_size(),
|
|
names);
|
|
}
|
|
p.end_map_processing(m_);
|
|
}
|
|
|
|
/*!
|
|
* \brief render a layer given a projection and scale.
|
|
*/
|
|
template <typename Processor>
|
|
void feature_style_processor<Processor>::apply_to_layer(layer const& lay,
|
|
Processor & p,
|
|
projection const& proj0,
|
|
double scale,
|
|
double scale_denom,
|
|
unsigned width,
|
|
unsigned height,
|
|
box2d<double> const& extent,
|
|
int buffer_size,
|
|
std::set<std::string>& names)
|
|
{
|
|
feature_style_context_map ctx_map;
|
|
layer_rendering_material mat(lay, proj0);
|
|
|
|
prepare_layer(mat,
|
|
ctx_map,
|
|
p,
|
|
scale,
|
|
scale_denom,
|
|
width,
|
|
height,
|
|
extent,
|
|
buffer_size,
|
|
names);
|
|
|
|
if (!mat.active_styles_.empty())
|
|
{
|
|
render_material(mat,p);
|
|
}
|
|
}
|
|
|
|
template <typename Processor>
|
|
void feature_style_processor<Processor>::prepare_layer(layer_rendering_material & mat,
|
|
feature_style_context_map & ctx_map,
|
|
Processor & p,
|
|
double scale,
|
|
double scale_denom,
|
|
unsigned width,
|
|
unsigned height,
|
|
box2d<double> const& extent,
|
|
int buffer_size,
|
|
std::set<std::string>& names)
|
|
{
|
|
layer const& lay = mat.lay_;
|
|
|
|
std::vector<std::string> const& style_names = lay.styles();
|
|
|
|
unsigned int num_styles = style_names.size();
|
|
if (num_styles == 0)
|
|
{
|
|
MAPNIK_LOG_DEBUG(feature_style_processor)
|
|
<< "feature_style_processor: No style for layer=" << lay.name();
|
|
return;
|
|
}
|
|
|
|
mapnik::datasource_ptr ds = lay.datasource();
|
|
if (!ds)
|
|
{
|
|
MAPNIK_LOG_DEBUG(feature_style_processor)
|
|
<< "feature_style_processor: No datasource for layer=" << lay.name();
|
|
return;
|
|
}
|
|
|
|
processor_context_ptr current_ctx = ds->get_context(ctx_map);
|
|
proj_transform prj_trans(mat.proj0_,mat.proj1_);
|
|
|
|
box2d<double> query_ext = extent; // unbuffered
|
|
box2d<double> buffered_query_ext(query_ext); // buffered
|
|
|
|
double buffer_padding = 2.0 * scale;
|
|
boost::optional<int> layer_buffer_size = lay.buffer_size();
|
|
if (layer_buffer_size) // if layer overrides buffer size, use this value to compute buffered extent
|
|
{
|
|
buffer_padding *= *layer_buffer_size;
|
|
}
|
|
else
|
|
{
|
|
buffer_padding *= buffer_size;
|
|
}
|
|
buffered_query_ext.width(query_ext.width() + buffer_padding);
|
|
buffered_query_ext.height(query_ext.height() + buffer_padding);
|
|
|
|
// clip buffered extent by maximum extent, if supplied
|
|
boost::optional<box2d<double> > const& maximum_extent = m_.maximum_extent();
|
|
if (maximum_extent)
|
|
{
|
|
buffered_query_ext.clip(*maximum_extent);
|
|
}
|
|
|
|
box2d<double> layer_ext = lay.envelope();
|
|
bool fw_success = false;
|
|
bool early_return = false;
|
|
|
|
// first, try intersection of map extent forward projected into layer srs
|
|
if (prj_trans.forward(buffered_query_ext, PROJ_ENVELOPE_POINTS) && buffered_query_ext.intersects(layer_ext))
|
|
{
|
|
fw_success = true;
|
|
layer_ext.clip(buffered_query_ext);
|
|
}
|
|
// if no intersection and projections are also equal, early return
|
|
else if (prj_trans.equal())
|
|
{
|
|
early_return = true;
|
|
}
|
|
// next try intersection of layer extent back projected into map srs
|
|
else if (prj_trans.backward(layer_ext, PROJ_ENVELOPE_POINTS) && buffered_query_ext.intersects(layer_ext))
|
|
{
|
|
layer_ext.clip(buffered_query_ext);
|
|
// forward project layer extent back into native projection
|
|
if (! prj_trans.forward(layer_ext, PROJ_ENVELOPE_POINTS))
|
|
{
|
|
MAPNIK_LOG_ERROR(feature_style_processor)
|
|
<< "feature_style_processor: Layer=" << lay.name()
|
|
<< " extent=" << layer_ext << " in map projection "
|
|
<< " did not reproject properly back to layer projection";
|
|
}
|
|
}
|
|
else
|
|
{
|
|
// if no intersection then nothing to do for layer
|
|
early_return = true;
|
|
}
|
|
|
|
std::vector<feature_type_style const*> & active_styles = mat.active_styles_;
|
|
|
|
if (early_return)
|
|
{
|
|
// check for styles needing compositing operations applied
|
|
// https://github.com/mapnik/mapnik/issues/1477
|
|
for (std::string const& style_name : style_names)
|
|
{
|
|
boost::optional<feature_type_style const&> style=m_.find_style(style_name);
|
|
if (!style)
|
|
{
|
|
continue;
|
|
}
|
|
|
|
if (style->comp_op() || style->image_filters().size() > 0)
|
|
{
|
|
if (style->active(scale_denom))
|
|
{
|
|
// we'll have to handle compositing ops
|
|
active_styles.push_back(&(*style));
|
|
}
|
|
}
|
|
}
|
|
return;
|
|
}
|
|
|
|
// if we've got this far, now prepare the unbuffered extent
|
|
// which is used as a bbox for clipping geometries
|
|
if (maximum_extent)
|
|
{
|
|
query_ext.clip(*maximum_extent);
|
|
}
|
|
|
|
box2d<double> & layer_ext2 = mat.layer_ext2_;
|
|
|
|
layer_ext2 = lay.envelope();
|
|
if (fw_success)
|
|
{
|
|
if (prj_trans.forward(query_ext, PROJ_ENVELOPE_POINTS))
|
|
{
|
|
layer_ext2.clip(query_ext);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
if (prj_trans.backward(layer_ext2, PROJ_ENVELOPE_POINTS))
|
|
{
|
|
layer_ext2.clip(query_ext);
|
|
prj_trans.forward(layer_ext2, PROJ_ENVELOPE_POINTS);
|
|
}
|
|
}
|
|
|
|
std::vector<rule_cache> & rule_caches = mat.rule_caches_;
|
|
attribute_collector collector(names);
|
|
|
|
// iterate through all named styles collecting active styles and attribute names
|
|
for (std::string const& style_name : style_names)
|
|
{
|
|
boost::optional<feature_type_style const&> style=m_.find_style(style_name);
|
|
if (!style)
|
|
{
|
|
MAPNIK_LOG_DEBUG(feature_style_processor)
|
|
<< "feature_style_processor: Style=" << style_name
|
|
<< " required for layer=" << lay.name() << " does not exist.";
|
|
|
|
continue;
|
|
}
|
|
|
|
std::vector<rule> const& rules = style->get_rules();
|
|
bool active_rules = false;
|
|
rule_cache rc;
|
|
for(rule const& r : rules)
|
|
{
|
|
if (r.active(scale_denom))
|
|
{
|
|
rc.add_rule(r);
|
|
active_rules = true;
|
|
collector(r);
|
|
}
|
|
}
|
|
if (active_rules)
|
|
{
|
|
rule_caches.push_back(std::move(rc));
|
|
active_styles.push_back(&(*style));
|
|
}
|
|
}
|
|
|
|
// Don't even try to do more work if there are no active styles.
|
|
if (active_styles.empty())
|
|
{
|
|
return;
|
|
}
|
|
|
|
double qw = query_ext.width()>0 ? query_ext.width() : 1;
|
|
double qh = query_ext.height()>0 ? query_ext.height() : 1;
|
|
query::resolution_type res(width/qw,
|
|
height/qh);
|
|
|
|
query q(layer_ext,res,scale_denom,extent);
|
|
q.set_variables(p.variables());
|
|
|
|
if (p.attribute_collection_policy() == COLLECT_ALL)
|
|
{
|
|
layer_descriptor lay_desc = ds->get_descriptor();
|
|
for (attribute_descriptor const& desc : lay_desc.get_descriptors())
|
|
{
|
|
q.add_property_name(desc.get_name());
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for (std::string const& name : names)
|
|
{
|
|
q.add_property_name(name);
|
|
}
|
|
}
|
|
q.set_filter_factor(collector.get_filter_factor());
|
|
|
|
// Also query the group by attribute
|
|
std::string const& group_by = lay.group_by();
|
|
if (!group_by.empty())
|
|
{
|
|
q.add_property_name(group_by);
|
|
}
|
|
|
|
bool cache_features = lay.cache_features() && active_styles.size() > 1;
|
|
|
|
std::vector<featureset_ptr> & featureset_ptr_list = mat.featureset_ptr_list_;
|
|
if (!group_by.empty() || cache_features)
|
|
{
|
|
featureset_ptr_list.push_back(ds->features_with_context(q,current_ctx));
|
|
}
|
|
else
|
|
{
|
|
for(std::size_t i = 0; i < active_styles.size(); ++i)
|
|
{
|
|
featureset_ptr_list.push_back(ds->features_with_context(q,current_ctx));
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
template <typename Processor>
|
|
void feature_style_processor<Processor>::render_material(layer_rendering_material & mat,
|
|
Processor & p )
|
|
{
|
|
std::vector<feature_type_style const*> & active_styles = mat.active_styles_;
|
|
std::vector<featureset_ptr> & featureset_ptr_list = mat.featureset_ptr_list_;
|
|
if (featureset_ptr_list.empty())
|
|
{
|
|
// The datasource wasn't queried because of early return
|
|
// but we have to apply compositing operations on styles
|
|
for (feature_type_style const* style : active_styles)
|
|
{
|
|
p.start_style_processing(*style);
|
|
p.end_style_processing(*style);
|
|
}
|
|
return;
|
|
}
|
|
|
|
p.start_layer_processing(mat.lay_, mat.layer_ext2_);
|
|
|
|
layer const& lay = mat.lay_;
|
|
|
|
std::vector<rule_cache> & rule_caches = mat.rule_caches_;
|
|
|
|
proj_transform prj_trans(mat.proj0_,mat.proj1_);
|
|
|
|
bool cache_features = lay.cache_features() && active_styles.size() > 1;
|
|
|
|
datasource_ptr ds = lay.datasource();
|
|
std::string group_by = lay.group_by();
|
|
|
|
// Render incrementally when the column that we group by changes value.
|
|
if (!group_by.empty())
|
|
{
|
|
featureset_ptr features = *featureset_ptr_list.begin();
|
|
if (features)
|
|
{
|
|
// Cache all features into the memory_datasource before rendering.
|
|
std::shared_ptr<featureset_buffer> cache = std::make_shared<featureset_buffer>();
|
|
feature_ptr feature, prev;
|
|
|
|
while ((feature = features->next()))
|
|
{
|
|
if (prev && prev->get(group_by) != feature->get(group_by))
|
|
{
|
|
// We're at a value boundary, so render what we have
|
|
// up to this point.
|
|
std::size_t i = 0;
|
|
for (feature_type_style const* style : active_styles)
|
|
{
|
|
|
|
cache->prepare();
|
|
render_style(p, style,
|
|
rule_caches[i],
|
|
cache,
|
|
prj_trans);
|
|
++i;
|
|
}
|
|
cache->clear();
|
|
}
|
|
cache->push(feature);
|
|
prev = feature;
|
|
}
|
|
|
|
std::size_t i = 0;
|
|
for (feature_type_style const* style : active_styles)
|
|
{
|
|
cache->prepare();
|
|
render_style(p, style, rule_caches[i], cache, prj_trans);
|
|
++i;
|
|
}
|
|
cache->clear();
|
|
}
|
|
}
|
|
else if (cache_features)
|
|
{
|
|
std::shared_ptr<featureset_buffer> cache = std::make_shared<featureset_buffer>();
|
|
featureset_ptr features = *featureset_ptr_list.begin();
|
|
if (features)
|
|
{
|
|
// Cache all features into the memory_datasource before rendering.
|
|
feature_ptr feature;
|
|
while ((feature = features->next()))
|
|
{
|
|
|
|
cache->push(feature);
|
|
}
|
|
}
|
|
std::size_t i = 0;
|
|
for (feature_type_style const* style : active_styles)
|
|
{
|
|
cache->prepare();
|
|
render_style(p, style,
|
|
rule_caches[i],
|
|
cache, prj_trans);
|
|
++i;
|
|
}
|
|
}
|
|
// We only have a single style and no grouping.
|
|
else
|
|
{
|
|
std::size_t i = 0;
|
|
std::vector<featureset_ptr>::iterator featuresets = featureset_ptr_list.begin();
|
|
for (feature_type_style const* style : active_styles)
|
|
{
|
|
featureset_ptr features = *featuresets++;
|
|
render_style(p, style,
|
|
rule_caches[i],
|
|
features,
|
|
prj_trans);
|
|
++i;
|
|
}
|
|
}
|
|
p.end_layer_processing(mat.lay_);
|
|
}
|
|
|
|
template <typename Processor>
|
|
void feature_style_processor<Processor>::render_style(
|
|
Processor & p,
|
|
feature_type_style const* style,
|
|
rule_cache const& rc,
|
|
featureset_ptr features,
|
|
proj_transform const& prj_trans)
|
|
{
|
|
p.start_style_processing(*style);
|
|
if (!features)
|
|
{
|
|
p.end_style_processing(*style);
|
|
return;
|
|
}
|
|
mapnik::attributes vars = p.variables();
|
|
feature_ptr feature;
|
|
bool was_painted = false;
|
|
while ((feature = features->next()))
|
|
{
|
|
bool do_else = true;
|
|
bool do_also = false;
|
|
for (rule const* r : rc.get_if_rules() )
|
|
{
|
|
expression_ptr const& expr = r->get_filter();
|
|
value_type result = util::apply_visitor(evaluate<feature_impl,value_type,attributes>(*feature,vars),*expr);
|
|
if (result.to_bool())
|
|
{
|
|
was_painted = true;
|
|
do_else=false;
|
|
do_also=true;
|
|
rule::symbolizers const& symbols = r->get_symbolizers();
|
|
if(!p.process(symbols,*feature,prj_trans))
|
|
{
|
|
for (symbolizer const& sym : symbols)
|
|
{
|
|
util::apply_visitor(symbolizer_dispatch<Processor>(p,*feature,prj_trans),sym);
|
|
}
|
|
}
|
|
if (style->get_filter_mode() == FILTER_FIRST)
|
|
{
|
|
// Stop iterating over rules and proceed with next feature.
|
|
do_also=false;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
if (do_else)
|
|
{
|
|
for( rule const* r : rc.get_else_rules() )
|
|
{
|
|
was_painted = true;
|
|
rule::symbolizers const& symbols = r->get_symbolizers();
|
|
if(!p.process(symbols,*feature,prj_trans))
|
|
{
|
|
for (symbolizer const& sym : symbols)
|
|
{
|
|
util::apply_visitor(symbolizer_dispatch<Processor>(p,*feature,prj_trans),sym);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
if (do_also)
|
|
{
|
|
for( rule const* r : rc.get_also_rules() )
|
|
{
|
|
was_painted = true;
|
|
rule::symbolizers const& symbols = r->get_symbolizers();
|
|
if(!p.process(symbols,*feature,prj_trans))
|
|
{
|
|
for (symbolizer const& sym : symbols)
|
|
{
|
|
util::apply_visitor(symbolizer_dispatch<Processor>(p,*feature,prj_trans),sym);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
p.painted(p.painted() | was_painted);
|
|
p.end_style_processing(*style);
|
|
}
|
|
|
|
}
|