Use pyproj new projection initialization method when setting a CRS in GeoPandas

Question

Issue

Using GeoPandas:

this:

(...)
t_srs = 4326
gdf.set_geometry('geom', crs={'init': u'epsg:'+str(t_srs)}, inplace=True)

or this:

(...)
gpd.read_postgis(sql_string, conn, geom_col='geom', crs={'init': u'epsg:'+str(t_srs)})

raise a warning today:

/usr/local/lib/python3.6/dist-packages/pyproj/crs.py:77:     
FutureWarning: 
    '+init=<authority>:<code>' syntax is deprecated.     
    '<authority>:<code>'  is the preferred initialization method.
  return _prepare_from_string(" ".join(pjargs))

and if I remove the 'init' key of the crs dictionary, I got this error;

CRSError: Invalid CRS input: {'epsg:4326'}

And if, by chance, I try with {'epsg':'4326'} or {'epsg':4326} ; these errors are respectively raised:

CRSError: Invalid projection: {"epsg": "4326"}: (Internal Proj Error: proj_create: Missing "type" key)

and:

CRSError: Invalid projection: {"epsg": 4326}: (Internal Proj Error: proj_create: Missing "type" key)

Question

Is it already possible to write some GeoPandas code that corresponds to the new* pyproj projection initialization method using <authority>:<code> instead of +init=<authority>:<code> ?

• https://pyproj4.github.io/pyproj/stable/gotchas.html

Answer 1

'<authority>:<code>' in the error means that you need just to use crs='EPSG:4326' instead of crs={'epsg:4326'}.

Answer 2

As too often, I continue cobbling around with my code after having posted my question, and I figure it out:

Short

Do not pass a dictionary but a string as a CRS!

Explanation

I.e, in concrete terms, change the two previous pieces of code that thrown the FutureWarning into

this:

(...)
t_srs = 4326
gdf.set_geometry('geom', crs=(u'epsg:'+str(t_srs)), inplace=True)

or this:

(...)
gpd.read_postgis(sql_string, conn, geom_col='geom', crs=(u'epsg:'+str(t_srs)))

Notice the parentheses instead of the curly braces, which mean now that crs is a string and no more a dictionary.

Here is, as an example, the help of the geopandas.read_postgis() method where it is discreetly suggested:

help(gpd.read_postgis)
Help on function read_postgis in module geopandas.io.sql:

read_postgis(sql, con, geom_col='geom', crs=None, index_col=None, coerce_float=True, parse_dates=None, params=None)
    Returns a GeoDataFrame corresponding to the result of the query
    string, which must contain a geometry column in WKB representation.

    Parameters
    ----------
    sql : string
        SQL query to execute in selecting entries from database, or name
        of the table to read from the database.
    con : DB connection object or SQLAlchemy engine
        Active connection to the database to query.
    geom_col : string, default 'geom'
        column name to convert to shapely geometries
    crs : dict or str, optional
        CRS to use for the returned GeoDataFrame; if not set, tries to
        determine CRS from the SRID associated with the first geometry in
        the database, and assigns that to all geometries.

    See the documentation for pandas.read_sql for further explanation
    of the following parameters:
    index_col, coerce_float, parse_dates, params

    Returns
    -------
    GeoDataFrame

    Example
    -------
    PostGIS
sql = "SELECT geom, kind FROM polygons"
    SpatiaLite
sql = "SELECT ST_AsBinary(geom) AS geom, kind FROM polygons"
df = geopandas.read_postgis(sql, con)