1. Modified Text Symbolizer
a) corrected line fragment centering (for 2nd and subsequent lines, when line breaks occur).
b) adjusted vertical alignment calculation so that:
i) middle -> has the center of the text line(s) at the point origin
ii) bottom -> has the text line(s) below the point origin
iii) top -> has the text line(s) above the point origin
c) added new text_symbolizer attribute: 'wrap_before', value range: true/false, default == false
allows line breaks at first wrap_char before wrap_width as an alternative to the original
which was to create the line break at the first wrap_char after wrap_width
d) added new text_symbolizer attribute: 'horizontal_alignment', value range: left/middle/right, default == middle
i) left -> has all text line(s) to left of the point origin
ii) middle -> has all text line(s) centered on the the point origin
iii) right -> has all text line(s) to the right of the point origin
NOTE: dx, dy position adjustments are applied after alignments and before Justify.
e) added new text_symbolizer attribute: 'justify_alignment', value range: left/middle/right, default == middle
i) left -> after alignments, has all text line(s) are left justified (left to right reading)
ii) middle -> after alignments, has all text line(s) center justified
iii) right -> after alignments, has all text line(s) right justified (right to left reading)
f) added new text_symbolizer attribute: 'opacity', value range: 0.0 thru 1.0; 1.0 == fully opaque
g) modified positioning to compensate for both line_spacing and character_spacing, to ensure proper
centering of the text envelope. Also ensure that centering occurs correctly even if no wrapping
occurs. Line spacing is uniform and consistent and compensates for errors between text_size and
the actual size (ci.height is inconsistent, depending on case and character); fixes issue with
multi-line text where some lines have a slight gap and others are compressed together.
2. Modified shield_symbolizer
a) added the attributes:
i) allow_overlap
ii) vertical_alignment
iii) horizontal_alignment
iv) justify_alignment
v) wrap_width
vi) wrap_character
vii) wrap_before
viii) text_convert
ix) line_spacing
x) character_spacing
xi) opacity
b) added new shield_symbolizer attribute: 'unlock_image', value range: true/false, default == false
i) false == image and text placement behaviour same as before
ii) true == image placement independant of text, image is always centered at geometry point, text placed per attributes,
dx/dy only affect text.
Allows user to create point markers with text, but both the text and image rendering collision detection are done
as a pair (they come and go together - solves problem if using point_symbolizer and text_symbolizers where one or the
other are omitted due to overlaps, but not both)
c) extended choices for the attribute 'placement' to include vertex; effect is limited to the shield_symbolizer
Allows an attempted placement at every vertex available, gives additional shield placement volume when using line geometry
d) ensured that the text placement was not updating the detector unless a shield image was actually placed.
e) added new shield_symbolizer attribute: 'no_text', value range: true/false, default = false
When set true, the text for the feature is ignored ('space' subsituted) so that pure graphic symbols can be used
and no text is rendered over top of them.
* text_convert="none|toupper|tolower"
Convert all text to upper/lower case before rendering. "none" doesn't do
anything with the text and is the default. Works for labels along lines
or at points.
* line_spacing="<number>"
Add this many pixels space between two lines in text labels that have
been broken into several lines. Default is 0. Doesn't do anything for
labels along lines.
* character_spacing="<number>"
Add this many pixels space between two characters in a text. Default is 0.
Currently only works for text labels on point geometries. This should
also be implemented for labels along lines, but I'll leave that for
another day.
* wrap_character="<character>"
Instead of breaking text into lines on spaces, use this character. This
is useful, when you want to make sure that labels are broken at the right
spot. Note that you'll probably want to make wrap_width small so that
your lines are actually broken, otherwise you'll see the wrap_character
in the output. Default is ' ' (space). Doesn't do anything for labels
along lines.
a) Raster opacity
b) Raster merging modes (TODO:add more modes, consider agg impl)
c) Raster scaling algos: fast,bilinear,bilinear8
(TODO: add alpha support in bilinear8)
d) improvements to png256
*Great work, thanks!*
- sqlite.input: added wkb_format parameter for selecting WKB format (generic/spatialite)
- sqlite.input: commented check of the spatial index
- wkb.hpp: removed wkqSQLite in favour of wkbSpatiaLite
replace it with support for rendering to a cairo context rather than
a cairo surface.
When rendering to a surface a show_page is done, when rendering to a
context it is not so that multiple renders can be done, possibly with
modified scaling and/or translation.
When you use save_xml() the colors are written out in #RGB format
which can not represent alpha information. The patch makes it save using
the rgb()/rgba() strings.
This adds the fontset support into save_xml
- Adds fontset_name into the text attributes
- Adds the fontset definitions
- glue to read fontsets from map
- build now osm plugin automatic if libxml2 is used
- moved osm/Makefile to osm/Makefile.example because if name collision
-> better move the example to another directory later
- optional CAIRO support
- changed libxml2 switch comment
transcode strings in filter expressions (default to utf8 input)
-- we can use map definition *.xml to define encoding
-- and in case of Python from
# -*- coding: utf-8 -*-
utf8 encoded filters are working now.
Add new find_point_placements to cover the gap left.
Change shield symbolizer to use find_point_placements.
Results are the same as before, but with much less duplicate code.
Slightly changes the result of max_char_angle_delta as now character angles aren't strict against the line angles.
Example screenshots on mailing list! :)
Spacing and no spacing line rendering are done by the same function.
Code is cleaner, performance should be equal or improved.
Text displacement is "Different", better in some cases and worse in others, I can revert it if wanted.
Old non spacing code is not yet removed as it is used by ShieldSymbolizer which I haven't investigated yet.
Individual character placement still has issues, that is my next task to fix!
1. The README talks about "scons.py", but it is just called "scons".
2. In demo/c++ there is a Makefile, but the readme.txt gives a command
line to compile. Both don't work for me. One error can at least be
fixed by adding -I../../agg/include to the command line.
3. When compiling Mapnik an error is thrown because of a redefined type.
This was on an Intel 64Bit machine. Fixed by "typedef char byte" to
"typedef uint8_t byte" in include/mapnik/global.hpp. I am not sure
whether thats the right fix, though, I just guessed. :-)
4. include/mapnik/image_data.hpp needs
#include <mapnik/octree.hpp>
to compile
Example:
>>> from mapnik import *
>>> im = Image(200,200)
>> save_to_file('test.png',im) # Guess image type from file extension
>> save_to_file('test.png,'jpeg',im) # Explicitly save as JPEG
2. PostGIS plug-in - optional 'multiple_geometries' parameter to control how Multi* geometries built.
3. MarkersSymbolizer (work in progress) to render vector shapes (markers) alonh a path with collision detection.
- libxml2 support
- strict error handling while parsing XML map files
- implemented save_map()
- removed some duplicate defaults
- all symbolizers with icons share a common base class now
(FIXME : label_spacing is still, too slow!!)
2. Re-use some agg objects.
3. placement_finder cleanups!
4. Added support for 'building_symbolizer' - extruded polygons
+ raster and gdal input isn't tested. Not working currently...
+ *-input.so plugins created. Change loader in source or link plugin to correct place
+ use pkg-config uninstalled feature
-> define project root to PKG_CONFIG_PATH to use mapnik without installation
+ added various library checks
+ don't install fonts
+ don't use included AGG
-> check for a installed libagg
+ Added Makefile for c++ demo
+ don't build any python wrapper stuff
-> this follows if all other building works
+ added Anjuta file
-> not needed to build anything, but helps much if you use Anjuta
Shield symbolizer is now a subclass of text symbolizer.
Some small improvements to text rendering.
Fixed up placement finder for horizontal placement.
Cleaned up placement finder.
2. Pass resolution to bbox query
3. Use variant<int,double,string> as parameter value e.g in Python:
ds = Raster(file="/path/to/file",lox = 12312.4,.....)
Added extractor facility to work with mapnik::parameter (C++):
mapnik::parameters params;
params["parameter0"] = 123.456;
params["parameter1"] = "123.456"; // initialize with string extract double later
boost::optional<double> val0 = params.get<double>("parameter0");
if (val0)
{
std::cout << *val0;
}
// with default value. NOTE: there is no 'parameter2' in params
boost::optional<double> val2 = params.get<double>("parameter2",654.321);
std::cout << * val2;
//
4. Added Gdal factory method in __init__.py
ds = Gdal(file="/tmp/file.tiff")
>>> for f in m.query_point([ layer.name for layer in m.layers ].index('world'),51,0):
... print f
...
feature (
f_code:FA001
fac_id:193
id:3147
na2:UK
na3:E
nam:ENGLAND
tile_id:10
)
>>>
fs = m.query_map_point(x,y) # Map (screen) coordinates
for feature in fs:
print feature
TODO: provide interface to feature in Python, at the moment only __str__ implemented which dumps attributes
to be rendered. This allows for a large extent (larger than can be
rendered into a single image in memory) to be rendered out as tiles.
Since the full extent is used for the placement calculations text
crossing tile boundaries will be consistent.
This method is a little inefficient when a large number of labels need
placed, an improved method would be to cache these placements between
tiles, but the attached is a start.
c++ users should simple call the render method with a start X and Y
coordinate specified,
for (int TileX = 0;TileX < 5;++TileX)
{
for(int TileY = 0;TileY < 5; ++TileY)
{
int TileSize=250;
int StartX = TileX*TileSize;
int StartY = TileY*TileSize;
Image32 buf(TileSize,TileSize);
agg_renderer<Image32> ren(m,buf,StartX,StartY);
ren.apply();
char name[324];
sprintf(name,"tile_%d_%d.png",TileX,TileY);
ImageUtils::save_to_file(name,"png",buf);
}
}
python users should call render_tile_to_file
for y in range(tile_count_y):
for x in range(tile_count_x):
if not os.path.exists("tiles/%d/%d/" % (map_scale, y)):
os.makedirs("tiles/%d/%d/" % (map_scale, y))
render_tile_to_file(m, x*tile_size, y*tile_size, tile_size, tile_size,
'tiles/%d/%d/%d.png' % (map_scale,y,x), 'png')
bool hit_test(double x, double y, double tol);
2. added image_view(unsigned x, unsigned y, unsigned width, unsigned height)
allowing to select region from image data e.g (in Python):
im = Image(2048,2048)
view = im.view(0,0,256,256)
save_to_file(filename,type, view)
3. changed envelope method to return vy value in datasource classes
4. features_at_point impl for shape and postgis plug-ins
angle changes too much between characters it finds an alternate
placement), this is specified in the max_char_angle_delta property in
radians (feel free to change it to degrees).
It also improves the text placement around corners trying to minimise
the distance between the center of the character and the line on each
side. This is the major portion of the patch.
