[Benchmarking] Data sets and styling

[Andrea Aime]

Hi all,
sending this mail to detail a bit more the maps we're going to use
for the benchmarking, taking as reference what we already wrote down
on the wiki, here:
http://wiki.osgeo.org/wiki/Benchmarking_2009#Layers

texas_roads_unstyled
--------------------------------------------------------
This one should be pretty much like the other years I guess.
Shall we use the exact same test? Maybe not, the hardware won't
be comparable anyways and we might want to make the test
comparable with the other tests using texas_roads instead
(we'll need one that jumps between scales for the 
texas_roads_styled_scale, shall we use the same set of WMS
bboxes for all other benchmarks?)

texas_roads_styled
--------------------------------------------------------
The data structure of the texas_roads layer is as follows:

perf=# \d tiger_pg
                                  Tabella "public.tiger_pg"
  Colonna  |         Tipo          |                      Modificatori 

----------+-----------------------+--------------------------------------------------------
  gid      | integer               | not null default 
nextval('tiger_pg_gid_seq'::regclass)
  cfcc     | character varying(3)  |
  cfcc_1   | character varying(1)  |
  cfcc_2   | character varying(1)  |
  cfcc_3   | character varying(1)  |
  module   | character varying(8)  |
  tlid     | numeric               |
  name     | character varying(35) |
  length_m | numeric               |
  altname1 | character varying(32) |
  altname2 | character varying(32) |
  altname3 | character varying(30) |
  altname4 | character varying(28) |
  altname5 | character varying(27) |
  altname6 | character varying(32) |
  altname7 | character varying(32) |
  altname8 | character varying(32) |
  altname9 | character varying(32) |
  the_geom | geometry              |

The CFCC code can be used in order to classify the roads and
is detailed here:
http://proximityone.com/tgrcfcc.htm

The actual data we have contains only Axxx and Pxxx codes, with only
21880 P ones out of a 3138920 total. I'd say we ignore A and P and
concentrate only on cfcc_2 that contains the main classification of the
road?

select cfcc_2, count(*) from tiger_pg group by cfcc_2 order by cfcc_2;
  cfcc_2 |  count
--------+---------
  1      |   18334
  2      |   62516
  3      |   89197
  4      | 2722910
  5      |   16895
  6      |   52959
  7      |  176109

I guess we could depict 1 and 2 with a highway like simple (thick orange
line with gray borders), 3 as yellow line with gray borders, 4 as
solid thin dark gray line and 5,6,7 as light gray dashed line?
Did not actually tried that out, but how does the idea look like?

texas_roads_styled_scale
-----------------------------------------------------------
Same styling as the above, but make a 3 zoom level plan where
only 1, 2 and 3 are shown when zoomed out, a middle level
that shows also 4, and finally show everything?

texas_roads_styled_labeled
-----------------------------------------------------------
Based on texas_roads_styled, but we show a big label with
halo on the highways, a smaller label with a thin halo for
class 4, and no label at all for the rest?
Kind of follows the availability of labels, roads in class
between 1 and 3 are mostly labelled, only 2/3 of the class
4 have a label and the others are mostly un-labeled anyways

# select cfcc_2, count(*) from tiger_pg where name is null group by 
cfcc_2 order by cfcc_2;
  cfcc_2 | count
--------+--------
  1      |      3
  2      |      9
  3      |     44
  4      | 901898
  5      |  15172
  6      |  31853
  7      | 154489

point_layer_styled
-----------------------------------------------------------
Hmmm... haven't got a clue here.
Would be nice to have a point layer in which a mix of
different icons can be used, and maybe do two maps out
of it, one using only simple vector symbols (square,
circle, triangle) and then one using external png or external
svg.

Does anyone have a nice map we can use?
Tiger does provide landmark point files but they are county based,
I guess one could schedule a download of all the Texas ones
and then merge them into a single postgis table, but that
will take a bit of time.

Any other suggestion? Maybe OSM data has point layers that
we can use for a classified point map?

polygon layers
----------------------------------------------------------
Not sure what to use here. Maybe a map of conties with
population count so that we can make a thematic map out
of it?
I accept suggestions.

Well, enough already for this mail. Feedback and suggestions
expected :)

Cheers
Andrea