c4py

Pure Python implementation of C4 universal content identification (SMPTE ST 2114:2017).

import c4py

# Verify a delivery matches the manifest
manifest = c4py.load("delivery-v3.c4m")
for path, entry in manifest.flat_entries():
    if entry.c4id and not c4py.verify(f"/deliveries/v3/{path}", entry.c4id):
        print(f"MISMATCH: {path}")

# Identify any file — same content always produces the same ID
c4id = c4py.identify_file("render.1001.exr")

# Compare two snapshots of a project
old = c4py.load("delivery-v2.c4m")
new = c4py.load("delivery-v3.c4m")
diff = c4py.diff(old, new)
print(f"+{len(diff.added)} -{len(diff.removed)} ~{len(diff.modified)}")

Install

pip install c4py

What is C4?

C4 IDs are universally unique, unforgeable identifiers derived from content using SHA-512. They are standardized as SMPTE ST 2114:2017. Same content always produces the same 90-character ID, regardless of filename, location, or time.

>>> import c4py
>>> c4py.identify_bytes(b"hello world")
C4ID('c41yP4cqy7jmaRDzC2bmcGNZkuQb3VdftMk6YH7ynQ2Qw4zktKsyA9fk52xghNQNAdkpF9iFmFkKh2bNVG4kDWhsok')

C4M Format

A c4m file is a human-readable text file that describes a filesystem. It captures file names, sizes, permissions, timestamps, and C4 IDs in a format you can read, edit, diff, and email.

-rw-r--r-- 2025-06-15T12:00:00Z      3 README.md   c45xZeXwMSpq...
drwxr-xr-x 2025-06-15T12:00:00Z      3 src/        -
  -rw-r--r-- 2025-06-15T12:00:00Z    3 main.go     c45KgBYEvEE7...

A 2 KB c4m file can describe an 8 TB project. Compare two c4m files to find exactly which frames changed across a delivery — in seconds, not hours.

API

Identification

# Identify a file on disk
c4id = c4py.identify_file("render.1001.exr")

# Verify a file matches an expected ID
assert c4py.verify("render.1001.exr", expected_id)

# From file-like object (streaming, constant memory)
c4id = c4py.identify(file_obj)

# From bytes
c4id = c4py.identify_bytes(b"data")

# Parse a C4 ID string (also works as C4ID constructor)
c4id = c4py.parse("c45xZeXwMSpq...")
c4id = c4py.C4ID("c45xZeXwMSpq...")  # same thing

# C4 ID properties
str(c4id)       # 90-character string
bytes(c4id)     # 64-byte digest
c4id.hex()      # hex digest
bool(c4id)      # False for nil ID, True otherwise

Tree IDs (Set Identity)

# Compute a single ID for a set of IDs (order-independent)
tree_id = c4py.tree_id([id_a, id_b, id_c])

C4M Files

# Parse
manifest = c4py.load("project.c4m")
manifest = c4py.loads(text)

# Write
c4py.dump(manifest, file_obj)
text = c4py.dumps(manifest)
text = c4py.dumps(manifest, pretty=True)

# Scan directory
manifest = c4py.scan("/path/to/dir")

# Iterate entries
for entry in manifest:
    print(entry.name, entry.size, entry.c4id)

Content Store

c4py shares the same content store as the c4 CLI and c4sh. Content stored by any tool is immediately available to the others.

# Open a store (auto-discovers from C4_STORE env or ~/.c4/config)
store = c4py.open_store()

# Or specify a path
store = c4py.open_store("/data/c4store")

# Store and retrieve content
c4id = store.put(open("render.exr", "rb"))
assert store.has(c4id)
content = store.get(c4id)

# Scan + store in one pass (zero extra I/O)
manifest = c4py.scan("/projects/HERO", store=store)

Diff and Patch Chains

# Compare two manifests
diff = c4py.diff(old_manifest, new_manifest)
diff.added      # entries only in new
diff.removed    # entries only in old
diff.modified   # same path, different content
diff.same       # identical entries

# Produce a c4m patch (like `c4 diff before.c4m after.c4m`)
patch_text = c4py.patch_diff(old_manifest, new_manifest)

# c4m files can contain version histories (base + patches)
# Resolve a chain to its final state (like `c4 patch project.c4m`)
final = c4py.resolve_chain(manifest)

# Enumerate patches (like `c4 log project.c4m`)
for info in c4py.log_chain(manifest):
    print(f"{info.index}  {info.c4id}  +{info.added} -{info.removed} ~{info.modified}")

Validation

result = c4py.validate(manifest)
for issue in result.errors:
    print(issue)

Works With

c4py is part of the C4 ecosystem:

• c4 — Go CLI for identification and content storage (c4 id, c4 cat, c4 diff)
• c4sh — Shell integration that makes c4m files behave as directories
• c4git — Git clean/smudge filter for large media assets

All tools share the same content store and produce identical C4 IDs.

Compatibility

c4py produces byte-identical output to the Go reference implementation. Cross-language test vectors ensure this.

Zero external dependencies. Pure Python. Works offline.

Design Decisions

See the FAQ for design decisions including SHA-512 permanence, the c4m format, and content store scaling.

License

Apache 2.0