Quick Reference
Purpose
This skill documents the 17 most frequent Sverchok errors: 7 common error patterns that cause silent data corruption, crashes, or wasted computation, and 10 AI-specific mistakes that LLMs consistently make when generating Sverchok code.
Critical Warnings
NEVER output flat data to sv_set() — ALWAYS wrap in the object-level list: sv_set([[data]]), not sv_set([data]).
NEVER mutate data retrieved with sv_get(deepcopy=False) — this corrupts the upstream node's cached output and all downstream consumers.
NEVER use __init__ for node initialization — ALWAYS use sv_init(self, context).
NEVER create sockets inside process() — sockets are created in sv_init() or sv_update() only.
NEVER use zip() directly on mismatched Sverchok lists — ALWAYS use match_long_repeat() first.
NEVER store IFC entity STEP IDs between IfcSverchok tree runs — SvIfcStore.purge() invalidates all IDs on every re-run.
ALWAYS use update=updateNode on every bpy.props property in a Sverchok node.
ALWAYS check output.is_linked before computing expensive results in process().
Decision Tree
Node produces wrong geometry silently?
├── Vertices treated as separate objects → Missing object wrapper: use [[(v1),(v2)]] not [(v1),(v2)]
├── Data appears truncated → Using zip() instead of match_long_repeat()
├── Upstream data corrupted after run → Mutating sv_get(deepcopy=False) data in-place
└── Numbers wrong but no error → Wrong nesting level for data type
Node property changes have no effect?
├── No update= callback on bpy.props → Add update=updateNode
├── Using custom lambda callback → Replace with updateNode
└── Tree.sv_process is False → Enable processing on the tree
Node errors on initialization?
├── Using __init__ instead of sv_init → Use sv_init(self, context)
├── Creating sockets in process() → Move socket creation to sv_init()
└── Wrong socket type alias in SNLite → Use single-letter aliases (v, s, m)
IfcSverchok-specific issues?
├── Entity IDs change between runs → SvIfcStore.purge() resets everything; query by GUID
├── Crash on undo → Known issue; save frequently, use "Re-run all nodes" instead
├── Data not reaching IFC nodes → Use double-nested lists: [["value"]] not ["value"]
└── Entities lost after re-run → Purge clears all entities; rebuild is by design
Part A: 7 Common Error Patterns
Error 1: Nesting Level Errors (Silent Data Corruption)
Wrong nesting level causes downstream nodes to misinterpret data structure. Each vertex is treated as a separate object instead of part of one object.
# WRONG: Flat vertex list: each vertex becomes a separate "object"
self.outputs['Vertices'].sv_set([(0,0,0), (1,0,0)])
# CORRECT: Object-wrapped vertex list: one object with 2 vertices
self.outputs['Vertices'].sv_set([[(0,0,0), (1,0,0)]])
Rule: ALWAYS wrap output data in the object-level list. Even a single object must be [[...]]. See sverchok-syntax-data for full nesting rules.
Error 2: Missing updateNode Callback
Property changes do not trigger node re-evaluation. The node appears stuck on its initial value.
# WRONG: No update callback: property changes are ignored
my_prop: FloatProperty(name='Scale', default=1.0)
# CORRECT: updateNode dispatches PropertyEvent to the update system
from sverchok.data_structure import updateNode
my_prop: FloatProperty(name='Scale', default=1.0, update=updateNode)
Error 3: Not Checking Output Connections
Unnecessary computation runs when outputs are not connected, wasting CPU time.
# WRONG: Always computes, even when no downstream node reads the output
def process(self):
data = expensive_computation()
self.outputs['Result'].sv_set(data)
# CORRECT: Early exit when no output is connected
def process(self):
if not any(s.is_linked for s in self.outputs):
return
data = expensive_computation()
self.outputs['Result'].sv_set(data)
Error 4: Socket Data Mutation (Upstream Corruption)
Modifying input data in-place corrupts the upstream node's cached output. All other downstream nodes reading from the same source receive corrupted data.
# WRONG: Mutating sv_get() data modifies the shared cache
data = self.inputs['Data'].sv_get(deepcopy=False)
data[0].append(999) # CORRUPTS upstream node's cached output
# CORRECT: Use deepcopy=True (default) when mutating
data = self.inputs['Data'].sv_get(deepcopy=True)
data[0].append(999) # Safe — working on an independent copy
# CORRECT: Or create new data without mutating (deepcopy=False for performance)
data = self.inputs['Data'].sv_get(deepcopy=False)
result = [sublist + [999] for sublist in data] # New list, original untouched
Error 5: List Matching Misunderstandings
Using Python's zip() silently truncates data to the shortest input. Sverchok's convention is to repeat the last element of shorter lists.
# WRONG: zip() truncates to shortest: data is silently lost
for v, s in zip(verts, scale):
...
# CORRECT: match_long_repeat extends shorter lists by repeating last element
from sverchok.data_structure import match_long_repeat
verts, scale = match_long_repeat([verts, scale])
for v, s in zip(verts, scale):
...
Error 6: IfcSverchok Undo Crashes
Blender crashes when undoing operations in IfcSverchok node trees. This is a known limitation of the IfcSverchok integration.
Workaround: Save frequently. Avoid undo in IfcSverchok trees. Use the "Re-run all nodes" button to restore state instead. See sverchok-impl-ifcsverchok for details.
Error 7: IfcSverchok Purge on Re-run
SvIfcStore.purge() is called before every full tree update, clearing all entity data. IFC entities are recreated from scratch on every update — STEP IDs change between runs.
Consequence: Code that stores IFC entity IDs will break on re-run. Query entities by GUID or type instead of by STEP ID. See sverchok-impl-ifcsverchok for entity management patterns.
Part B: 10 AI Common Mistakes
AI Mistake 1: Flat Vertex Data
AI generates flat number lists instead of the required nested structure.
# AI MISTAKE: Returning flat list: downstream nodes fail or misinterpret
result = [1, 2, 3]
self.outputs['Numbers'].sv_set(result) # WRONG
# CORRECT: Always wrap in object level
self.outputs['Numbers'].sv_set([[1, 2, 3]])
AI Mistake 2: Missing Object Wrapper for Vertices
AI omits the outer list that represents the object level, causing each vertex to be treated as a separate object.
# AI MISTAKE: Vertices at wrong nesting level
verts = [(0,0,0), (1,0,0), (1,1,0)]
self.outputs['Vertices'].sv_set(verts) # WRONG — each vertex = separate object
# CORRECT: Vertices at level 3 (object list > vertex list > coordinate tuple)
self.outputs['Vertices'].sv_set([[(0,0,0), (1,0,0), (1,1,0)]])
AI Mistake 3: __init__ Instead of sv_init
AI uses Python's standard __init__ instead of Sverchok's sv_init lifecycle method. Blender node classes do not support __init__ for socket creation.
# AI MISTAKE: __init__ is NOT how Sverchok nodes initialize
def __init__(self):
self.inputs.new(...) # WRONG — will fail or be ignored
# CORRECT: Use sv_init: called once when node is created
def sv_init(self, context):
self.inputs.new('SvStringsSocket', 'Data')
self.outputs.new('SvStringsSocket', 'Result')
AI Mistake 4: Wrong SNLite Socket Type Aliases
AI invents socket type names instead of using Sverchok's single-letter aliases.
# AI MISTAKE: Invalid type aliases in SNLite header
"""
in verts vertices # WRONG — "vertices" is not a valid alias
out result string # WRONG — "string" is not a valid alias
"""
# CORRECT: Use single-letter aliases
"""
in verts v # v = SvVerticesSocket
out result s # s = SvStringsSocket
"""
# Valid aliases: s=Strings, v=Vertices, m=Matrix, c=Color
AI Mistake 5: Forgetting updateNode on Properties
AI defines bpy.props properties without the update=updateNode callback, causing property changes to have no visible effect.
# AI MISTAKE: Property changes are silently ignored
count: IntProperty(name='Count', default=5) # WRONG
# CORRECT: Always include updateNode
from sverchok.data_structure import updateNode
count: IntProperty(name='Count', default=5, update=updateNode)
AI Mistake 6: zip() Instead of match_long_repeat
AI uses Python's zip() which silently truncates mismatched lists. Sverchok's design expects the last element to be repeated.
# AI MISTAKE: zip() silently drops data from longer list
for v, s in zip(verts, scale): # WRONG — truncates to shortest
# CORRECT: match_long_repeat extends shorter lists
from sverchok.data_structure import match_long_repeat
verts, scale = match_long_repeat([verts, scale])
for v, s in zip(verts, scale):
...
AI Mistake 7: Creating Sockets in process()
AI creates sockets inside process(), which runs on every update. This adds duplicate sockets on each evaluation.
# AI MISTAKE: Creates new socket on EVERY update cycle
def process(self):
self.inputs.new('SvStringsSocket', 'Extra') # WRONG — duplicates accumulate
# CORRECT: Create sockets in sv_init() (once) or sv_update() (on topology change)
def sv_init(self, context):
self.inputs.new('SvStringsSocket', 'Extra')
AI Mistake 8: Assuming Matrix Data Is Nested Like Vertices
AI wraps matrices in double-nested lists like vertices. Matrices use a different nesting convention — they are at the object level directly.
# AI MISTAKE: Treating matrices like vertices (extra nesting level)
matrices = [[Matrix(), Matrix()]] # WRONG — over-nested
# CORRECT: Matrices are at object level (level 1), not nested further
matrices = [Matrix(), Matrix()]
self.outputs['Matrix'].sv_set(matrices)
AI Mistake 9: IfcSverchok Entity ID Persistence
AI stores IFC entity STEP IDs and assumes they persist across tree re-runs. SvIfcStore.purge() resets all data before each full update.
# AI MISTAKE: Caching STEP IDs between runs
saved_id = 42 # WRONG — this ID is invalid after next tree update
# CORRECT: Query by GlobalId (GUID) or by type, never by STEP ID
# GUIDs are stable across re-runs; STEP IDs are not
wall = ifc_file.by_guid("2O2Fr$t4X7Z8jBCV$0Oo9a")
See sverchok-impl-ifcsverchok for stable entity reference patterns.
AI Mistake 10: IfcSverchok Single vs Double Nesting
AI sends single-nested data to IfcSverchok nodes, which expect standard Sverchok double-nested lists.
# AI MISTAKE: Single-nested data to IFC nodes
self.outputs['Names'].sv_set(["Wall_001"]) # WRONG — flat list
# CORRECT: IfcSverchok nodes expect standard Sverchok nesting
self.outputs['Names'].sv_set([["Wall_001"]])
Error Detection Checklist
Use this checklist when debugging Sverchok node issues:
| Symptom | Likely Error | Fix |
|---|---|---|
| Each vertex rendered as separate object | Missing object wrapper (Error 1, AI 2) | Wrap in [[...]] |
| Property slider has no effect | Missing updateNode (Error 2, AI 5) |
Add update=updateNode |
| Data silently truncated | Using zip() (Error 5, AI 6) |
Use match_long_repeat() |
| Upstream data changes unexpectedly | Socket mutation (Error 4) | Use deepcopy=True |
| Sockets multiply on each run | Creating sockets in process() (AI 7) |
Move to sv_init() |
| Node init error | Using __init__ (AI 3) |
Use sv_init(self, context) |
| SNLite type error | Wrong socket alias (AI 4) | Use v, s, m, c |
| Matrix transform broken | Wrong nesting (AI 8) | Use [M1, M2] not [[M1, M2]] |
| IFC entities disappear on re-run | Purge resets (Error 7, AI 9) | Query by GUID, not STEP ID |
| IFC node receives no data | Single nesting (AI 10) | Use [["value"]] |
| Blender crash on undo in IFC tree | IfcSverchok undo bug (Error 6) | Save often, avoid undo |
| Slow node tree builds | Not using init_tree() |
Wrap in with tree.init_tree(): |
Cross-References
- sverchok-syntax-data — Full data nesting rules, socket types, and nesting level specifications per data type
- sverchok-impl-ifcsverchok — IfcSverchok node patterns, SvIfcStore lifecycle, entity management
- sverchok-impl-custom-nodes — Custom node lifecycle (sv_init, process, sv_update), registration, socket management
- sverchok-core-concepts — Node tree architecture, update system, event types
Reference Links
- references/methods.md — Error detection methods, diagnostic functions, debugging utilities
- references/examples.md — Complete working code examples demonstrating correct patterns for each error
- references/anti-patterns.md — Detailed anti-patterns with WHY explanations and fix strategies