Hi Michel,

Thanks! I think I got the suggestions by Anna, Pietro and Moritz working, but this certainly looks like an handy function to have at hand. One possible disadvantage if using this in a script to be shared with others is that it would add an dependency on pandas, wouldn't it?

Cheers,

Paulo


On 01-10-15 10:01, Michel Wortmann wrote:
Hi Paulo,
I had the same problem some time ago. I am using pandas for a lot of things and discovered the quickest way to append a column to a vector table is using pandas' .to_sql dataframe method. Here is my full function for it:

import pandas as pa
dataframe = pa.DataFrame( ... )
def appendColumns(vecttbl,dataframe,join=None,layer=1):
     '''Upload a DataFrame (Series should be first converted into a sql 
compliant datafram)
     to a grass vector table. Join is a valid column in vecttbl or if None take 
categories.
     Always joins on df.index, which is not uploaded.'''
     # get sqlite.db info
     db = grass.vector_db(vecttbl)[layer]
     con = sqlite3.connect(db['database'])
     # decide join indeces
     itbl = {False:db['key'],True:join}[join!=None]
     # make series to dataframe
     dataframe.to_sql('pandas_temp',con,if_exists='replace',index_label='ix') # 
index label?
     # join columns to vecttbl
     grass.run_command('v.db.join',map=vecttbl,column=itbl,
                        other_table='pandas_temp',other_column='ix',
                        subset_columns=','.join(map(str,dataframe.columns)))
     # remove temp table
     cur = con.cursor()
     cur.execute('DROP TABLE IF EXISTS pandas_temp')
     con.commit(); con.close()
     return

The dataframe.to_sql line is essentially all you need, the rest of the function is just prepping.
Hope it helps,
Michel


On 10/01/2015 12:05 AM, Paulo van Breugel wrote:


On 30-09-15 17:52, Paulo van Breugel wrote:


On 30-09-15 17:47, Anna Petrášová wrote:


On Wed, Sep 30, 2015 at 11:20 AM, Paulo van Breugel <p.vanbreu...@gmail.com> wrote:



    On 30-09-15 16:18, Anna Petrášová wrote:


    On Wed, Sep 30, 2015 at 9:27 AM, Paulo van Breugel
    <p.vanbreu...@gmail.com <mailto:p.vanbreu...@gmail.com>> wrote:



        On Wed, Sep 30, 2015 at 11:53 AM, Pietro
        <peter.z...@gmail.com> wrote:

            On Wed, Sep 30, 2015 at 9:51 AM, Paulo van Breugel
            <p.vanbreu...@gmail.com> wrote:
            >
            >
            > On Wed, Sep 30, 2015 at 2:02 AM, Anna Petrášová
            <kratocha...@gmail.com>
            > wrote:
            >>
            >>
            >>
            >> On Tue, Sep 29, 2015 at 6:09 PM, Paulo van Breugel
            >> <p.vanbreu...@gmail.com> wrote:
            >>>
            >>> This must be a very basic question, but I can't
            find an easy/direct way
            >>> to do this. In python, if I have an array with
            values with a length equal to
            >>> the number of rows in an attribute table of a
            (point) vector layer, how can
            >>> I write those values to a new column in that
            attribute table. I can of
            >>> course first create the column, but than how to
            update that column with the
            >>> values in the array?
            >>
            >>
            >> it should be pretty easy to do with pygrass,
            unfortunately there is no
            >> example on assigning attributes in the official
            documentation [1],  but it
            >> should be pretty easy, something like that (not
            tested):
            >>
            >> with VectorTopo('myvector', mode='w') as vectormap:
            >>     for feature in vectormap:
            >>  feature.attrs['mycolumn'] = value
            >>
            >>
            > Thanks, but that seems to write the vector back
            without attribute table

            You have to save the changes in the database out from
            your cycle, with:

            vectormap.table.conn.commit()


        Thanks Pietro. I am, however, not sure I understand (I
        tried to use it, but thanks to my limited experience in
        Python / pygrass not much luck). Just to be more specific,
        I am trying to create a script that divides points in
        training and test groups, similar to v.kcv, but with
        points clustered in space. E.g.,

        # Create vector
        grass.run_command("v.random", output="testB", npoints=10,
        overwrite=True)
        grass.run_command("v.db.addtable", map="testB", columns="X
        DOUBLE PRECISION,Y DOUBLE PRECISION,GR INTEGER")
        grass.run_command("v.to.db", map="test", option="coor",
        columns="X,Y")

        # Create groups
        vectmap = 'test'
        cvals = array(grass.vector_db_select(vectmap, layer =
        int(1), columns = 'X,Y')['values'].values()).astype(np.float)
        centroids,_ = kmeans(cvals,2)
        idx,_ = vq(cvals,centroids)

        # write results to tabel
        Now I would like to write idx to the column 'GR' in the
        attribute table of 'test'.

        p.s. I am first creating the XY columns now, but is there
        a function to get the coordinates (cvals) in pygrass directly?


    I don't fully understand the example,

    Thanks for the quick response. I basically have a list with
    values (idx in the example above) which I like to add as a
    column to the attribute table of an existing vector (point
    layer). The length of idx is equal to the number of rows in the
    attribute table. The solution of Anna seems like an elegant
    solution (and easier and more flexible than other solutions I
    tried using e.g., sqlite3). However, as I wrote, I end up with
    a vector without attribute table. You wrote that I "have to
    save the changes in the database out from your cycle, with:
    vectormap.table.conn.commit()". I am, however, not sure what
    you mean with 'out from your cycle' or how to implement that.


probably after the for cycle ends you would call this 'vectormap.table.conn.commit()', if it doesn't works, try to put it in the cycle (I am not sure what is supposed to work).

I did try both, with no luck so-far. I'll give it another try though, perhaps I did something else wrong.

Tried out again, but after running the code below, I end up with an empty vector layer (no points and no attribute table). Same if I put the vectormap.table.conn.commit() in the for cycle.

import grass.script as grass
from grass.pygrass.vector import VectorTopo
grass.run_command("v.random", output="testB", npoints=10, overwrite=True)
grass.run_command("v.db.addtable", map="testB", columns="GR INTEGER")
with VectorTopo('testB', mode='w') as vectormap:
     for feature in vectormap:
         feature.attrs['GR'] = 9
         vectormap.table.conn.commit()


    but yes, you can get coordinates:

    with VectorTopo('myvector', mode='w') as vectormap:
    for feature in vectormap:
      print feature.x
          print feature.y

    Great, thanks. I did not find this in the manual. If it is not
    there, perhaps it would be something worth including? I would
    not mind providing a text, but I am not sure what would be the
    best way to do that.


Technically, it's in the manual:
https://grass.osgeo.org/grass70/manuals/libpython/pygrass_vector.html#geometry-classes

but we are lacking more examples of often used constructions. Some other examples are available in the recent workshop we did:

https://github.com/wenzeslaus/python-grass-addon/blob/master/02_pygrass_library.ipynb

Thanks, I will have a look at it.




            Pietro









_______________________________________________
grass-dev mailing list
grass-dev@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-dev



_______________________________________________
grass-dev mailing list
grass-dev@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-dev

_______________________________________________
grass-dev mailing list
grass-dev@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-dev

Reply via email to