### Install nsqip-tools

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Installs the NSQIP Tools Python package using pip.

```bash
pip install nsqip-tools
```

--------------------------------

### Network Drive Support for Data Processing

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Illustrates the package's compatibility with network drives and file systems that lack file locking. It shows how to build Parquet datasets and load data from network locations seamlessly.

```python
# Works on network drives, SMB shares, etc.
result = nsqip_tools.build_parquet_dataset(
    data_dir="/Volumes/network_drive/nsqip_data",
    output_dir="/Volumes/network_drive/processed"
)

# Query from network location
query = nsqip_tools.load_data("/Volumes/network_drive/processed/adult_nsqip_parquet")
```

--------------------------------

### Build Parquet Dataset

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Builds an NSQIP parquet dataset from text files, applying standard transformations and optionally generating a data dictionary.

```python
import nsqip_tools

# Build parquet dataset from NSQIP text files
result = nsqip_tools.build_parquet_dataset(
    data_dir="/path/to/nsqip/files",
    dataset_type="adult"  # or "pediatric"
)

print(f"Dataset created at: {result['parquet_dir']}")
print(f"Data dictionary at: {result['dictionary']}")
```

--------------------------------

### Build Parquet Dataset

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Builds a Parquet dataset from specified data directories. Supports automatic memory detection or custom memory limit specification for efficient data processing and storage. Uses columnar storage for optimized compression and access.

```python
# Use automatic memory detection (default)
result = nsqip_tools.build_parquet_dataset(data_dir="/path/to/files")

# Or specify custom limit
result = nsqip_tools.build_parquet_dataset(
    data_dir="/path/to/files",
    memory_limit="8GB"
)
```

--------------------------------

### NSQIP Tools API Reference

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Provides detailed information on the functions available in the nsqip-tools package, including parameters, return values, and usage.

```APIDOC
build_parquet_dataset(data_dir, output_dir=None, dataset_type="adult", generate_dictionary=True, memory_limit="4GB", verify_case_counts=True, apply_transforms=True)
  Build an NSQIP parquet dataset from text files with standard transformations.
  Parameters:
    data_dir: Path to NSQIP text files
    output_dir: Output directory (defaults to data_dir)
    dataset_type: "adult" or "pediatric"
    generate_dictionary: Generate data dictionary
    memory_limit: Memory limit for operations
    verify_case_counts: Verify case counts match expected
    apply_transforms: Apply standard transformations
  Returns: Dictionary with paths to:
    parquet_dir: Parquet dataset directory
    dictionary: Data dictionary CSV file (if generated)
    log: Build log file

load_data(path_to_parquet_dataset)
  Load NSQIP data from a parquet dataset for querying.
  Returns: Query object for chaining filter and collection methods.

Query Object Methods:
  filter_by_cpt(cpt_codes)
    Filter by CPT procedure codes.
  filter_by_diagnosis(diagnosis_codes)
    Filter by ICD diagnosis codes.
  filter_by_year(years)
    Filter by operation years.
  filter_active_variables()
    Keep only variables with data in most recent year.
  select_demographics()
    Select common demographic variables.
  select_outcomes()
    Select common outcome variables.
  lazy_frame
    Get the Polars LazyFrame for custom operations.
  collect()
    Execute query and return Polars DataFrame.
  count()
    Get count of rows without collecting full data.
  sample(n)
    Get a random sample of n rows.
  describe()
    Get summary statistics about the query.
```

--------------------------------

### Safe Data Collection and Sampling

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Demonstrates memory-safe data collection practices, including checking dataset size before collection, using streaming for large datasets, and obtaining samples for exploration. This prevents out-of-memory errors.

```python
# Check size before collecting
query = nsqip_tools.load_data("/path/to/parquet/dataset").filter_by_year([2021])
info = query.describe()
print(f"Total rows: {info['total_rows']}")
print(f"Columns: {info['columns']}")

# Use streaming for large datasets
df = query.collect(streaming=True)

# Get a sample for exploration
sample_df = query.sample(n=10000)
```

--------------------------------

### Load and Query Data

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Loads NSQIP data from a parquet dataset and demonstrates filtering by CPT codes, years, and diagnosis codes using chained Polars operations.

```python
import nsqip_tools
import polars as pl

# Load and filter data
df = (
    nsqip_tools.load_data("/path/to/parquet/dataset")
    .filter_by_cpt(["44970", "44979"])  # Laparoscopic procedures
    .filter_by_year([2020, 2021])
    .collect()
)

# Chain with Polars operations
df = (
    nsqip_tools.load_data("/path/to/parquet/dataset")
    .filter_by_diagnosis(["K80.20"])  # Gallstones
    .lazy_frame  # Access the Polars LazyFrame
    .select(["CASEID", "AGE_AS_INT", "CPT", "OPERYR"])
    .filter(pl.col("AGE_AS_INT") > 50)
    .group_by("CPT")
    .agg(pl.count())
    .collect()
)
```

--------------------------------

### System Memory Information

Source: https://github.com/brant01/nsqip_tools/blob/main/README.md

Retrieves and displays system memory details, including total RAM, available memory, and a recommended memory limit for processing. This is useful for understanding system capacity before running memory-intensive operations.

```python
import nsqip_tools

# Check system memory
mem_info = nsqip_tools.get_memory_info()
print(f"Total RAM: {mem_info['total']}")
print(f"Available: {mem_info['available']}")
print(f"Recommended limit: {mem_info['recommended_limit']}")
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.