tommy's scratchpad

I’ve been helping Paul with setting up the OSGeo Toronto Code Sprint 2009 event here in Toronto in March.

This promises to be an effective and fun event, to have many developers from the various OSGeo projects in one place for a few days.

Check it out, and hope to see you there!

Check out this video if you’re interested in learning more about the evolution of open source.  Pretty neat!

Check out “Open Source Approaches in Spatial Data Handling” by Hall, Leahy et. al. (disclosure: I did chapter 1).  An interesting read covering many facets and tools of open source geospatial.

I’ve been researching tag clouds in the last few days. I think tag clouds can help geospatial search front ends in giving the user a “weighted list”, to get them to what they want quickly and more efficiently.

tag cloud example

The following Python script takes a list of terms as input.  Such a list can be derived from many things, such as an existing taxonomy, analyzing an httpd log file for commonly used search terms, user votes, and so on.  In this (simple) example, we use comma separated input.

By creating a term and count dictionary, this sets up the anatomy of a tag cloud.  From here, you can pass this for output to the web (i.e. font sizes, colours, etc.).  Here we output this to an APML document, which is often used to represent tag clouds.  You can then use tools such as cluztr to generate tag clouds with ease.

Considerations:

• the script does a very simple job to assign values of 0.0 to 1.0 to weights
• It would be neat to apply these to searches against spatial identifiers (i.e. “Montreal”), and then map them accordingly
• It would be interesting to hear Cartographers’ thoughts on the tag cloud concept

#!/usr/bin/python

import sys
import fileinput
import datetime
from lxml import etree

# term dictionary
dTags = {}
tn = datetime.datetime.now().isoformat()

for line in fileinput.input(sys.argv[1]):
    aTags = line.strip().split(",")
    for sTag in aTags:
        # if term is not in list, add
        if sTag not in dTags:
            dTags[sTag] = 1
        # else increment term count
        else:
            dTags[sTag] += 1 

# output as APML document
node = etree.Element('APML', nsmap={None: 'http://www.apml.org/apml-0.6'})
node.attrib['version'] = '0.6'
subnode = etree.Element('Body')
subnode.attrib['defaultprofile'] = 'owscat'
subsubnode = etree.Element('Profile')
subsubnode.attrib['defaultprofile'] = 'Terms'
subsubsubnode = etree.Element('ImplicitData')
subsubsubsubnode = etree.Element('Concepts')

for term, count in sorted(dTags.iteritems()):
    termnode = etree.Element('Concept')
    termnode.attrib['key']     = term
    termnode.attrib['value']   = str(float(float(count/10.0)))
    termnode.attrib['from']    = 'owscat'
    termnode.attrib['updated'] = str(tn)
    subsubsubsubnode.append(termnode)

subsubsubnode.append(subsubsubsubnode)
subsubnode.append(subsubsubnode)
subnode.append(subsubnode)
node.append(subnode)

print etree.tostring(node, xml_declaration=True, encoding='UTF-8', pretty_print=True)

I’ve written my share of catalogues, Capabilities parsers, map clients, and context import/export tools to know that having good example WMS instances is paramount in testing functionality and building features. I usually have a handy list of WMS servers which I constantly use when writing code.

Bird Studies Canada provides WMS access to their various bird distribution and abundance data. BSC has taken every effort to:

• populate their Capabilities metadata exhaustively. Title, abstract, keywords, and even MetadataURL pointers to FGDC XML documents for all layers. And _full_ service provider metadata (including Attribution, which is great for displaying Logo images, etc.)
• return GetFeatureInfo in both GML and HTML for prettier responses

This WMS is always at the top of my testing list, as well as my first response when people ask to see an existing WMS example which is well constructed, and serves catalogues and search demos very well indeed.

Kudos to BSC!

I’m starting to work on contributing SOS and OWS Common support in OWSLib, a groovy and regimented little GIS Python project.

So far so good; some initial implementations are done (committing soon hopefully, writing tests around these).  I think this will add value to the project, seeing that SOS 1.0 has been around long enough to start seeing implementations.  And the OWS Common support will act as a baseline for all calling specs/code to leverage.

And it’s been a nice journey in Python for me so far.  Another thing I like about this project is the commitment to testing — awesome!

A piece of work I help out with involves the visualization and access of hydrometric monitoring data over the Web. Part of this involves the data management and publishing of voluminous databases of monitoring information.

We use Chameleon for basic visualization and query of the data. Behind the scenes, we run a slew of complex processes (shell scripts via cron) to output the data in a format that can be understood by MapServer (which we use to publish WMS layers). The processes work across many disparate database connections, so outputting them to shapefiles and accessing them locally helps with performance in web mapping apps. ogr2ogr is used exclusively and extensively for the access and format translation.

Well, today I found out that an effort began to write a bunch of scripts to additionally output OGC KML. Thank goodness things didn’t get very far, because the following addition to our processes:

$ ogr2ogr -F KML foo.kml bar.ovf -dsco NameField=NAME -dsco DescriptionField=COMMENT

…worked like a charm, and put a big smile on people’s faces!

So now, OGC KML is also supported for visualization in Earth browsers. Just like that.

Output styles are relatively simple; I’m thinking a -dsco like:

-dsco LayerStyle=LayerName,styles.kml#mystyle

…would point to an existing (local or remote) KML style document style ID via XPointer, i.e.:

<styleUrl>somefile.kml#mystyle</styleUrl>

Of course the default behaviour would be in place if this -dsco is not defined. I’ll see what the GDAL KML gurus think about this.

At any rate, once again, thank you GDAL for being an uber-utility for day-to-day GIS tasks. Happy faces everywhere!

I needed to do some pre-processing of some data which involved transposing column names to values. The condition was that the value for each respective column (frequency count) had to be > 1.

My input was a csv file, and my goal was an output csv file which would feed into a batch database import process.

ID,DA,NL,PHENOM1,PHENOM2,PHENOM3,PHENOM4
233,99,44,0.00,27.00,12.00,0.00

The other interesting bit was that only a range of columns applied to the condition; the other columns represented ancillary data.

Enter Python:

#!/usr/bin/python

import sys
import csv

# open file and read headers
fPhenomenon = open("phenomenon.txt","r")
sHeaders    = fPhenomenon.readline().replace(r'"','')
aHeaders    = sHeaders.split(",")

# feed the rest to csv
csvIn  = csv.reader(fPhenomenon)
csvOut = csv.writer(sys.stdout)

for sRowIn in csvIn:
    aRowOut = []
    aPhenomenon = []
    aRowOut.append(sRowIn[0]) # procedure ID
    aRowOut.append(sRowIn[1]) # major drainage area ID
    for nIndexTupleVal, tupleVal in enumerate(sRowIn[3:-1]):
        if (float(tupleVal) > 0): # phenomenon measured at least once
            # add phenomenon name to list
            aPhenomenon.append(aHeaders[nIndexTupleVal+3])
        # add phenomenon list to record
        aRowOut.append(",".join(aPhenomenon))
    csvOut.writerow(aRowOut)

Notes

• hooray for raw strings!
• enumerate() is great and saves you the trouble of declaring your own counter
• like any language, modules/libraries makes things so easy to work with
• I wish the header stuff was a bit cleaner (I should look further into the csv module w.r.t. headers

That’s my hack for the day. Have a good weekend!

UPDATE: ah, the csv module has a .next() method, which can be used instead of the shoemaker attempt I made above to regularize / split / store the header list.

Inspired by the recent thread on FOSS4G history, I started an effort to document MapServer’s history, from its beginnings in the mid-1990s. Check out the progress we’ve made so far. If there’s anything missing, or in error, feel free to contribute!

Mateusz posted a link to an interesting topic on osgeo-discuss. I think it’s a great idea to document the history of geospatial and open source, and I echo Dave’s comments on how Wikipedia would be an ideal home for documentation and maintenance.

Perhaps the best way to go about this would be for the various projects on Wikipedia (MapServer, GDAL, GeoTools, GRASS, etc.) to document their respective histories, and allow the main Wikipedia OSGeo page to link to them accordingly.

Thoughts? Are there better alternatives than Wikipedia? Should projects document history on their own respective websites, which Wikipedia then references