bpy_data_materials

Cycles Materials (Nodes, no Nodes)
Internal Materials (Nodes, no Nodes)

Cycles Materials - scripted

Node Groups

Perhaps you want to manually define materials and node groups, and import ('append') them from another .blend via a script.

todo

Nodes

Scripting a Cycles node tree is quite easy. The only thing I find moderately annoying about it is that if I want the node tree to be user-ready (spread out a little) then I must pass the locations of the nodes explicitly. If you don't pass locations all nodes are added to (x=0, y=0). You might not care, and if you don't expect any user interaction with the nodes then you certainly don't need to add locations.

todo

NodeArrange to the rescue

Luckily there's a small addon by JuhaW called NodeArrange that is able to spread the nodes out in a tree form.

import bpy

import addon_utils
addon_utils.enable("NodeArrange-master")

mat = bpy.data.materials.new('Demo2')
mat.use_nodes = True
nodes = mat.node_tree.nodes 

# remove default diffuse node, get output node
nodes.remove(nodes.get('Diffuse BSDF'))
material_out = nodes.get('Material Output')

bsdf1 = nodes.new(type='ShaderNodeBsdfDiffuse')
bsdf1.inputs[0].default_value = 0, 0.2, 0.9, 1.0

bsdf2 = nodes.new(type='ShaderNodeBsdfDiffuse')
bsdf2.inputs[0].default_value = 0.8, 0.2, 0.9, 1.0

value1 = nodes.new(type='ShaderNodeValue')
mixer1 = nodes.new(type='ShaderNodeMixShader')

links = mat.node_tree.links
links.new(value1.outputs[0], mixer1.inputs[0])
links.new(bsdf1.outputs[0], mixer1.inputs[1])
links.new(bsdf2.outputs[0], mixer1.inputs[2])
links.new(mixer1.outputs[0], material_out.inputs[0])

bpy.ops.node.arrange_nodetree(mat_name=material_name)

# -- or even --
# bpy.ops.node.arrange_nodetree(
#   mat_name=material_name, margin_x=140, margin_y=120
# )

No Nodes