Count street length of a city

For a precise result, you cannot hope to cut out "all streets within a city" from OSM before you compute the length, because streets can cross the city boundary and if I understand you correctly, in this case you'd only want the length of the bit inside.

One way of getting this done is:

1. Import OSM data for a rectangle around your city into a PostGIS database with osm2pgsql. This will not only give you the roads but also the city boundary if mapped in OSM.
2. Use PostGIS to sum up the length of geometries inside your area of interest ("clipping polygon") as discussed in "Actual road length of exported map"

QGIS will also allow you to filter out all roads and clip them at the polygon boundary but it is a more cumbersome and manual process.

Yet another possiblity: you can run a query on Overpass API. This one produces a sample statistic for my home town, by value of highway. Length is in meters.

This should also point you to the real issues: You should decide for yourself whether you want to count footways and tracks or not. Another source of incertainty are dual carriageways; they would be counted twice because they are two ways in OpenStreetMap. Frontage roads, turn lanes and more may also affect the results. Finally, the request keeps ways on the boundary intact, thus slightly overestimating length values for that part of the problem.

Nontheless, for best effort guess data is for sure good enough.

Have you tried this package?

https://github.com/JoaoCarabetta/osm-road-length

It is super easy,

Install it: pip install osm-road-length

And run:

import osm_road_length from shapely import wkt

geometry = wkt.loads('POLYGON((-43.2958811591311 -22.853167273541693,-43.30961406928735 -23.035275736044728,-43.115980036084224 -23.02010939749927,-43.157178766552974 -22.832917893834313,-43.2958811591311 -22.853167273541693))')

length = osm_road_length.get(geometry)