mef3io

A single C++ core for MEF 3.0 read/write, wrapped for Python (nanobind) and MATLAB (MEX). The high-level semantics of the legacy mef_tools — float scaling, NaN discontinuities, precision inference, the int32-counts + conversion-factor primitive write path — live in C++, so every language binding behaves identically.

Documentation: https://bnelair.github.io/mef3io/ — install, Python / MATLAB / C++ guides, the MEF 3.0 format reference, and measured comparisons against the legacy stack.

Status

Read and write are implemented and cross-validated against the legacy pymef / mef_tools stack in both directions (~112 Python tests + standalone C++ tests). Scope notes:

Video (.vidd/.vmet/.vidx) is out of scope (traversal skips it).
Encryption is none or fully-encrypted (section 2 → L1 key, section 3 → L2 key); "level-1 only" is not a valid MEF file — see docs/encryption_model.md. Unlike the legacy reader, an L1 password actually works (signal + technical metadata, subject metadata locked) — see docs/legacy_comparison.md.
Records: read/write Note, EDFA, SyLg, Seiz (others read as header-only).
Pure-Python backend: not yet implemented (backend="pure" raises).
Append extends the channel's last segment in place (legacy semantics); new_segment=True forces a fresh segment.

Full docs live at https://bnelair.github.io/mef3io/ (built from docs/ with MkDocs Material, deployed by .github/workflows/docs.yml). In-repo: METADATA.md (session/subject/acquisition metadata), docs/mef3_format.md (format reference), docs/design.md (design), docs/legacy_comparison.md (measured performance

encryption comparison), benchmarks/README.md, runnable examples/, and CLAUDE.md.

Install

Python — prebuilt wheels (no compiler needed) for Linux (x86_64/aarch64), macOS (arm64/x86_64), and Windows (AMD64/ARM64), Python 3.10+:

pip install mef3io                 # runtime (numpy only)
pip install "mef3io[test]"         # + oracle tests (pymef, mef-tools, pandas, pytest)
pip install "mef3io[bench]"        # + NWB-Zarr benchmark stack

MATLAB — download the mef3io-matlab-vX.Y.Z.zip bundle from the GitHub release (classes + CI-tested MEX for all platforms, one addpath), or compile once on your machine (needs a C++20 compiler configured via mex -setup C++; ~30 s). See matlab/README.md:

run('<repo>/matlab/build_mex.m'); addpath('<repo>/matlab')
r = mef3io.Reader('session.mefd'); x = r.read('ch1');

From source (Python development) — needs CMake ≥ 3.26 and a C++20 compiler; uses the active env's Python:

scripts/dev_build.sh            # builds build/dev, symlinks the extension into python/mef3io
python -m pytest tests          # conftest sets up paths; needs pymef/mef-tools as oracle
python -m build --wheel         # build a wheel

All bindings share one version: the repo-root VERSION file feeds the wheel metadata, mef3io::version() in C++, and the MEX build.

The legacy mef-tools package (PyPI; import name mef_tools) is the correctness oracle and benchmark baseline: the test/benchmark scripts use a local mef_tools checkout when run inside the original repo, and otherwise the pip-installed mef-tools — so a standalone checkout just needs pip install "mef3io[test]" (or [bench]).

Usage

import numpy as np, mef3io

# Write. NaN = discontinuity gap; precision inferred if omitted.
with mef3io.Writer("session.mefd", overwrite=True, units="uV") as w:
    w.write("eeg1", float_signal, start_uutc, fs=256.0)
    # amplifier counts + V/bit, stored verbatim:
    w.write_int32("eeg2", counts_int32, ufact=2.5e-7, start_uutc=start_uutc, fs=256.0)
    w.write_annotations([{"time": start_uutc, "text": "note"}], channel="eeg1")

# Read. Partial reads by uUTC; gaps are NaN; scaled by the conversion factor.
with mef3io.Reader("session.mefd", n_threads=0) as r:   # 0 = all cores
    r.channels
    r.info("eeg1")
    x  = r.read("eeg1", t0, t1)       # float64, NaN gaps
    xi = r.read_raw("eeg1", t0, t1)   # {samples: int32, valid: mask, ...}
    r.segments("eeg1")                # per-segment map: what data is where
    r.toc("eeg1")                     # block index for viewers/seeking
    r.records("eeg1")                 # annotations

Warm-start metadata cache (opt-in)

Off by default. cache="auto" snapshots channel metadata into the per-user OS cache dir (safe for read-only sessions); a path makes it persistent. Warm opens serve channels/info without touching the tree and defer the backend until a data read. Invalidated automatically on any size/mtime change or on write.

mef3io.Reader("session.mefd", cache="auto")          # temp/OS cache dir
mef3io.Reader("session.mefd", cache="/data/s.cache") # persistent

Drop-in for existing pipelines

# from mef_tools.io import MefReader, MefWriter
from mef3io import MefReader, MefWriter

Same call shapes and semantics as the legacy classes (NaN-gap splitting, precision inference, int32 arrays via the primitive path, in-segment appends). See examples/07 and docs/legacy_comparison.md.

Parallelism

RED block decode and encode run on an internal thread pool (n_threads: 0 = hardware concurrency, 1 = serial; per-call override on reads). Output is byte-identical regardless of thread count. Bindings release the GIL, so external dataloader parallelism is safe too.

Layout

core/        C++17/20 library (types, byteio, crc, crypto, headers, metadata,
             red, session, reader, records, writer, session_writer) + Catch2 tests
bindings/    nanobind extension (_mef3io)
python/mef3io/  Reader, Writer, compat (mef_tools.io shim), cache, pure (stub)
matlab/      MEX gateway over the C ABI (core/include/mef3io/c_api.h),
             +mef3io Reader/Writer classes, build_mex.m, test_mef3io.m
examples/    runnable scripts: write/read, int32, append, segment map,
             annotations, encryption, legacy drop-in, replicability checks
tests/       golden fixture generator + P1–P9 pytest suites (pymef oracle)
benchmarks/  bindings + legacy/NWB-Zarr comparison scripts
docs/        MkDocs site source (guides, format reference, legacy comparison)
scripts/     dev_build.sh (dev build + extension symlink)

Contributing & support

Bug reports and feature requests via issues, questions via discussions. See CONTRIBUTING.md to build and test locally, and SECURITY.md to report vulnerabilities privately. Please cite via CITATION.cff.

License

Apache 2.0. Developed at the Mayo Clinic Bioelectronics Neurophysiology and Engineering Lab (BNEL).

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mef3io-1.1.2.tar.gz (414.9 kB view details)

Uploaded Aug 5, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

mef3io-1.1.2-cp314-cp314-win_arm64.whl (248.0 kB view details)

Uploaded Aug 5, 2026 CPython 3.14Windows ARM64

mef3io-1.1.2-cp314-cp314-win_amd64.whl (262.8 kB view details)

Uploaded Aug 5, 2026 CPython 3.14Windows x86-64

mef3io-1.1.2-cp314-cp314-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (407.9 kB view details)

Uploaded Aug 5, 2026 CPython 3.14manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mef3io-1.1.2-cp314-cp314-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (355.8 kB view details)

Uploaded Aug 5, 2026 CPython 3.14manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

mef3io-1.1.2-cp314-cp314-macosx_11_0_arm64.whl (233.0 kB view details)

Uploaded Aug 5, 2026 CPython 3.14macOS 11.0+ ARM64

mef3io-1.1.2-cp314-cp314-macosx_10_15_x86_64.whl (253.3 kB view details)

Uploaded Aug 5, 2026 CPython 3.14macOS 10.15+ x86-64

mef3io-1.1.2-cp313-cp313-win_arm64.whl (241.0 kB view details)

Uploaded Aug 5, 2026 CPython 3.13Windows ARM64

mef3io-1.1.2-cp313-cp313-win_amd64.whl (256.1 kB view details)

Uploaded Aug 5, 2026 CPython 3.13Windows x86-64

mef3io-1.1.2-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (407.8 kB view details)

Uploaded Aug 5, 2026 CPython 3.13manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mef3io-1.1.2-cp313-cp313-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (355.5 kB view details)

Uploaded Aug 5, 2026 CPython 3.13manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

mef3io-1.1.2-cp313-cp313-macosx_11_0_arm64.whl (232.8 kB view details)

Uploaded Aug 5, 2026 CPython 3.13macOS 11.0+ ARM64

mef3io-1.1.2-cp313-cp313-macosx_10_15_x86_64.whl (253.4 kB view details)

Uploaded Aug 5, 2026 CPython 3.13macOS 10.15+ x86-64

mef3io-1.1.2-cp312-cp312-win_arm64.whl (241.1 kB view details)

Uploaded Aug 5, 2026 CPython 3.12Windows ARM64

mef3io-1.1.2-cp312-cp312-win_amd64.whl (256.2 kB view details)

Uploaded Aug 5, 2026 CPython 3.12Windows x86-64

mef3io-1.1.2-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (407.8 kB view details)

Uploaded Aug 5, 2026 CPython 3.12manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mef3io-1.1.2-cp312-cp312-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (355.6 kB view details)

Uploaded Aug 5, 2026 CPython 3.12manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

mef3io-1.1.2-cp312-cp312-macosx_11_0_arm64.whl (232.9 kB view details)

Uploaded Aug 5, 2026 CPython 3.12macOS 11.0+ ARM64

mef3io-1.1.2-cp312-cp312-macosx_10_15_x86_64.whl (253.3 kB view details)

Uploaded Aug 5, 2026 CPython 3.12macOS 10.15+ x86-64

mef3io-1.1.2-cp311-cp311-win_arm64.whl (241.8 kB view details)

Uploaded Aug 5, 2026 CPython 3.11Windows ARM64

mef3io-1.1.2-cp311-cp311-win_amd64.whl (257.0 kB view details)

Uploaded Aug 5, 2026 CPython 3.11Windows x86-64

mef3io-1.1.2-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (409.2 kB view details)

Uploaded Aug 5, 2026 CPython 3.11manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mef3io-1.1.2-cp311-cp311-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (356.9 kB view details)

Uploaded Aug 5, 2026 CPython 3.11manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

mef3io-1.1.2-cp311-cp311-macosx_11_0_arm64.whl (233.9 kB view details)

Uploaded Aug 5, 2026 CPython 3.11macOS 11.0+ ARM64

mef3io-1.1.2-cp311-cp311-macosx_10_15_x86_64.whl (254.2 kB view details)

Uploaded Aug 5, 2026 CPython 3.11macOS 10.15+ x86-64

mef3io-1.1.2-cp310-cp310-win_amd64.whl (257.2 kB view details)

Uploaded Aug 5, 2026 CPython 3.10Windows x86-64

mef3io-1.1.2-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (409.6 kB view details)

Uploaded Aug 5, 2026 CPython 3.10manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mef3io-1.1.2-cp310-cp310-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (357.1 kB view details)

Uploaded Aug 5, 2026 CPython 3.10manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

mef3io-1.1.2-cp310-cp310-macosx_11_0_arm64.whl (234.2 kB view details)

Uploaded Aug 5, 2026 CPython 3.10macOS 11.0+ ARM64

mef3io-1.1.2-cp310-cp310-macosx_10_15_x86_64.whl (254.5 kB view details)

Uploaded Aug 5, 2026 CPython 3.10macOS 10.15+ x86-64

File details

Details for the file mef3io-1.1.2.tar.gz.

File metadata

Download URL: mef3io-1.1.2.tar.gz
Upload date: Aug 5, 2026
Size: 414.9 kB
Tags: Source
Uploaded using Trusted Publishing? No
Uploaded via: twine/7.0.0 CPython/3.12.13

File hashes

Hashes for mef3io-1.1.2.tar.gz
Algorithm	Hash digest
SHA256	`8848732e9ffcff3ab88b25298ea69da72353858f803b777e68484fbe1c5b210a`
MD5	`eae78f7f786a9a51ba43998ece835dfe`
BLAKE2b-256	`3706cb6e2137ca0885b32636b35219fefd10d1b4ac08d1af5bfc3f61c03ff047`

Algorithm	Hash digest
SHA256	`26b6855626c76efaf8923ab70afa251c8e3de74856eae694dcd0d09d25a68a22`
MD5	`ab482a2630b68a5f6ddc343e1c2c3dc8`
BLAKE2b-256	`5cd6db9bed62d05903bd77926fe5d5add8eea92da9ca29630682b468d7e15fa3`

Algorithm	Hash digest
SHA256	`b37a30ed7900741adb3570ef949972c5b8a9c2a380366e120ce862a1606a15bd`
MD5	`5b0c14ad5744afcdbb5322a387a4831f`
BLAKE2b-256	`58269e975fa4e824a5bcbfce1f692e476b0a7b6c83e6695196eeca98916fcbd3`

Algorithm	Hash digest
SHA256	`bb7dfd223d499bab9676445ae9763a20eee42d57aae4b3bf594bc169bad00dbb`
MD5	`561081bdbf082db6094782009ee1e125`
BLAKE2b-256	`dfb693d9d740025f2062dbe975dd1b27a043168759ce462a80147752e1c6941b`

Algorithm	Hash digest
SHA256	`6e9dd0d3e287cda75b348f3592e76edefaa0ade6b3ad0f32d321b3cea921f350`
MD5	`c07906cb3037189b19a0a40ed16ddc38`
BLAKE2b-256	`12c80c789e99bfd7f223ba3073c5009e78caefd0b2605eb0a54fa1b790cd8ef7`

Algorithm	Hash digest
SHA256	`45484e48096fbdbc9eede4649b80b2eeb44ff566b3a9abcf652f6ea70a34c70c`
MD5	`b40b650f11bc10fa05cb6efb93bee947`
BLAKE2b-256	`d806bffcb9c2c8c2ca49ab57dc84ddf826ecbaca7bb9f3c32ae02adc42c7d1a0`

Algorithm	Hash digest
SHA256	`6d6bb4132d8e2895b93335527401658bd54e003aad9e462d3db536a779042a98`
MD5	`15baeae717c03501620fad9f8a9edc0a`
BLAKE2b-256	`80cdd37ee5ca66e12e7959492b7a1366dc9c1920588ae1dab8a606802b2bec82`

Algorithm	Hash digest
SHA256	`3100fb61950786fd61789dde8ba80a686183408fe5dbcb2d76ae291d7c1c1f99`
MD5	`e0f1861f7eea5bf4121c0f8432ac308c`
BLAKE2b-256	`701d160293f913542d27d13ddcd9a9902c41ae93d91a51abd2b2d0f65e9f8d42`

Algorithm	Hash digest
SHA256	`b647cb32bab509fe0a85f306e3dcf37ff944edae1c3d604d644726c5035e48fa`
MD5	`0a854d82bac6f804a4d60e5cc909c792`
BLAKE2b-256	`e7371bd28ae6168d3fe88e7f90ead816300789986cbf814166559584c687301e`

Algorithm	Hash digest
SHA256	`f670a77e35fbf02968af44ee8523f404e311c0bc4d01958ae48c104cabb50bd0`
MD5	`37b28ecd283dbbe9f79f05cee578ac1d`
BLAKE2b-256	`a735a0865e49a17dc8ad60f1a4e4c3b107948d9a0ed83b6c558e3fc217710d20`

Algorithm	Hash digest
SHA256	`90c83fd6b16ae27bb083fbaa6811f24bc0c7e08b47fd4a697e9b35caf053243a`
MD5	`d5310ee0474cb885559b32a6d2a7b7a2`
BLAKE2b-256	`a27b8b319b2fc042181662b99ec4ad809e733d38596e48bea1c6b2265aa050d9`

Algorithm	Hash digest
SHA256	`30dbd8183e9ccc01e294e826408d9521d7b9fbe982403b29da5310405641d7f9`
MD5	`93e47d18d78a5aa089f4eb50f86eb971`
BLAKE2b-256	`96889f979e33f93a9f5f86c82cfc175780d72226ac4a807007b139923c8e678f`

Algorithm	Hash digest
SHA256	`2947894ebac66bba8f8d1ea8992ebe15ecc71fffe3b7566c5fadd93dc9340014`
MD5	`d1e9149ca43990235c4541a47d158e2f`
BLAKE2b-256	`da8dc690605f8930ec2f4628804e6edc9c2ff88a69ae4d0c85b65e5622a5ffa2`

Algorithm	Hash digest
SHA256	`da80b6f1d6984eee5a2a75eac9a9df0fd3ad670246dd8ddd55a39fdc25af8a74`
MD5	`faac886dfaa5a1fd6b031888a2c50b0e`
BLAKE2b-256	`bd9c38414ff8e47b06fd1200e670bd6c42944466b192e7ed5c4d153140c2e341`

Algorithm	Hash digest
SHA256	`e435130b2b752eea20220da5f6071c228c211e94ba024a5a81ac9940a6bfc95d`
MD5	`be99811a2ed5ea7a8791e64be0d8065c`
BLAKE2b-256	`756dd48711cb2f6ff506fe2919347a2330d5a9c39da1f9c0b4a5b1a1e14b7374`

Algorithm	Hash digest
SHA256	`55d43d00ddadfca0040c64d352e7766cdc8d9be08ce0fe63293a29c788a07565`
MD5	`b44b9df8ab0fdddeb38806091ace5764`
BLAKE2b-256	`ae1f7f0cca112790452c7194f9425ed64b968a88228478d1775dfaa9c252559c`

Algorithm	Hash digest
SHA256	`d2ecd2acd47ab5981d4cad2663f6e71900fd2b6707189b9fec94d937e4b0a55a`
MD5	`bfe4a9c47f67e89e0d952fe29b2b8db2`
BLAKE2b-256	`62d03a3dc7feafa87ba312cd17958d29942692f7632bf6b4c877a77a7823945c`

Algorithm	Hash digest
SHA256	`e759c8c3995b5e66cde3d8c58a1f6783464ac75375936df15c93319625b4056d`
MD5	`168a5683a80852af959f0d33a26fb3b7`
BLAKE2b-256	`c353f42724ff11acb1a3d1a493dc1918e2ddb521b060826e4d1dc7c95f34cbc0`

Algorithm	Hash digest
SHA256	`348bc910d7c99728a7c74f63b03e646f3fdcf9ffc9faba145767700b40a071ad`
MD5	`7c624383dd2f9320666de876dbaff54d`
BLAKE2b-256	`14ff2f9130eceaf737e772290cd9c3eca2b822b24d0c364ca04004a0614acd48`

Algorithm	Hash digest
SHA256	`ad7243185e9f50b76d9be020664c480e06a2576d34ebc8854c1caabffd8faede`
MD5	`06f91596d14236b847d03f34e94886a5`
BLAKE2b-256	`cf3e57213d1649d17779baffbbb302d10c13018878c6d6cb860c6e8c8d0a9c30`

Algorithm	Hash digest
SHA256	`15fcef7e0f03d2d87213072a699e87f0cea50020f66c9c0807410c3b729d01d4`
MD5	`0378c4894563eae7011f580b3c7f0390`
BLAKE2b-256	`5316267239d9114dd59d6b5c6bb66e3e1f3a9d3e64615ca7814303b53b53694b`

mef3io 1.1.2

mef3io

Status

Install

Usage

Warm-start metadata cache (opt-in)

Drop-in for existing pipelines

Parallelism

Layout

Contributing & support

License

Project links

Key dates

1 maintainer

Credits

License

Requires

Provides Extra

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

Algorithm	Hash digest
SHA256	`35208fb09fdc87f7916f532c7b330ea46ead9fbf8b8caf5b00c0542104f9638c`
MD5	`66aad2923ba83bb76e4bb0da883db4fa`
BLAKE2b-256	`f8004ccfb22ace2bd7d106384e2e98a01a2a99e62b460b8a51fe8c705a78332a`

Algorithm	Hash digest
SHA256	`03b70d1bd4d0eaaf617063354befefdab8b47232ec9b49dd8ab0ca281e593416`
MD5	`d9855d43cfda1fb43e9dcb630d2341e3`
BLAKE2b-256	`4fde7ad23d2de67793f48c0a5fcd4ab9211c1df7d76acb01d9764a13870224b0`

Algorithm	Hash digest
SHA256	`4b9d4036e37b4e0cd4b4abc0e0b6f09805ec40e07162f476055c1b3d13ae9a2b`
MD5	`2b60e642c051f9b08dfb4c7c8c05193f`
BLAKE2b-256	`48d87b1272122691e704298b6f2599909a61f86776392faa1113d5f2866d088a`

Algorithm	Hash digest
SHA256	`c5527a8984dda21847f59f14058e89526893945562d16416da9e5cd3077defd2`
MD5	`132a3ddf5369ce5dec20c86b2bddfcd8`
BLAKE2b-256	`f57357d56c87924a4795d2cf4976cca243825159a75737a0e76f596005135a95`

Algorithm	Hash digest
SHA256	`3e64a4d409091c1748380d156e613482be130cc8cde06c91323c428dd36b993f`
MD5	`e1cfbc7b6037d551a9e3303232d78e91`
BLAKE2b-256	`b65df76eda6e6b16d821ed6f306da829b0c9e2df74c15be1f8055f570e83f613`

Algorithm	Hash digest
SHA256	`ec571fc187083d40ef2b1203a69264f16986e717a99f2348aefc1c6208c5f86d`
MD5	`be29bd1d51f23f7f0bb59436424ac5b6`
BLAKE2b-256	`f4b88151d23ec686079aaad6071406f1b832b6b78928da5d4977563d3071314a`

Algorithm	Hash digest
SHA256	`b2f4aa401a02b1b33a5f365d8b9e43ada28af18c60173c5b05229af44f077cee`
MD5	`a1684aefc1d74fc52f40abc19012b2e4`
BLAKE2b-256	`035171f40f1d5399e14cd66faf240f3759a9f6e92dccfff695861e2938d46ebf`

Algorithm	Hash digest
SHA256	`fca03cbe98027f6a7e007e6c055c1846a867cebd808a7c5169c24e8c39144fd1`
MD5	`904da476d162539f9659e8e5c5eb9801`
BLAKE2b-256	`963d3a644221720eaf884dcff34eaf0b2c15f63fa53cc420491bb44237f9d12a`

Algorithm	Hash digest
SHA256	`e692ed97b527307e69cc229fea202ddba642707f75fc5c4ae8657ec5426df787`
MD5	`2c6dc46f4d2f39dc27f8ffc6f87dee1c`
BLAKE2b-256	`f9ef09c665f8be491c44eeb91fb15bcf0af6d4ee87ef0c36247bea22112b56b8`