Getting started `converter()` issue - ValueError issues

[NG-sama]

ValueError Issue occurs during conversion of .h5ad file to .slaf. Issue identified due to use of hardcoded "data" value as input for X_Group.

Issue output as follows

File ".venv/lib/python3.13/site-packages/slaf/data/converter.py", line 2361, in _validate_optimized_dtypes
    if "data" in X_group:

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Solution:

Within converter.py perform the following changes (commented lines are changes):

            # Check the original data type from the h5ad file
            X_group = reader.file["X"]
            data = X_group.get("data") if hasattr(X_group, "get") else None

            # if "data" in X_group:
            if data is not None:
                # Sample some original data
                # data = X_group["data"]
                sample_size = min(10000, len(data))
                # Use sequential sampling instead of random to avoid indexing issues
                sample_data = data[:sample_size]

Code as follows:

import marimo as mo

from slaf import SLAFArray
from slaf.integrations import scanpy as slaf_scanpy
from slaf.integrations.anndata import read_slaf
import scanpy as sc
from slaf.data.converter import SLAFConverter

dtest = sc.datasets.pbmc3k_processed()
# Convert to SLAF format
converter = SLAFConverter()
slaf_path = "pbmc3k.slaf"
converter.convert("/filepath/pbmc3k_processed.h5ad",slaf_path)

print(f"✅ Converted to SLAF format: {slaf_path}")

[NG-sama]

Creating a slaf dataset with this method does not create all the cell metadata columns

7 out of 10 metadata columns only created (as per the 01-getting-started.py page)

Output created by slaf.info()

2026-03-08 00:58:18.727 | INFO     | slaf.core.slaf:display_ascii_art:40 -
  z Z
 ( - . - )
 /  ^ \ ⚡
(  (_)  (_) )
 \_______/
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:393 -
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:418 - 📊 SLAF Dataset Loaded: pbmc3k.slaf
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:419 -    • Shape: 2,638 cells × 1,838 genes
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:420 -    • Format: SLAF v0.4
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:428 -    • Expression records: 4,848,644
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:432 -    • Sparsity: 0.0%
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:438 -    • Optimizations: use_integer_keys: True, optimize_storage: True
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:441 -    • Status: Ready for queries (metadata loading in background)
2026-03-08 00:58:18.728 | INFO     | slaf.core.slaf:_display_initialization_message:442 -

✅ Loaded SLAF dataset: 2,638 cells × 1,838 genes
SLAF Dataset
  Shape: 2638 cells × 1838 genes
  Format version: 0.4
  Cell metadata columns: 7
    n_genes, percent_mito, n_counts, louvain, cell_id...
  Gene metadata columns: 3
    n_cells, gene_id, gene_integer_id
  Record counts:
    Cells: 2,638
    Genes: 1,838
    Expression records: 4,848,644
    Sparsity: 0.0%
    Density: 100.0%
  Expression statistics:
    Min value: 0.000
    Max value: 65535.000
    Mean value: 358.832
    Std value: 0.000
  Optimizations:
    use_integer_keys: True
    optimize_storage: True

[pavanramkumar]

Thanks for reporting! For the first bug, can you share which dataset (if not the default pbmc3k) and which version of slafdb you're using? I pushed a fix for a similar bug recently (see #30). Can you update your slafdb installation and try in case you haven't already

[pavanramkumar]

I'm not able to repro the second issue on my end but I can be more useful if you report the source h5ad file used in your test.

Edit: I tried this and saw only 4 columns in the original source file

import scanpy as sc
d = sc.datasets.pbmc3k_processed()
print("Number of obs columns:", len(d.obs.columns))
print("Obs columns:", list(d.obs.columns))

[NG-sama]

Thanks Pavan.

For Issue#1 converter() issue - updating slafdb to v0.4.0 resolved the issue.

For issue#2 - can you provide me which scanpy dataset is converted in the getting-started.py notebook? I was assuming that sc.datasets.pbmc3k_processed() was the one used.