Parsing HEC-RAS Compute Messages

Primary Sources (navigate to these for complete details):

• HDF Class Reference:

  ras_commander/hdf/AGENTS.md

  - Class hierarchy, decorators
• HdfResultsPlan Implementation:

  ras_commander/hdf/HdfResultsPlan.py

  - Compute message methods
• Working Example:

  examples/400_1d_hdf_data_extraction.ipynb

  - Compute message extraction

This skill provides patterns for extracting and interpreting HEC-RAS computation output. For implementation details, see the primary sources above.

Quick Start

Check Plan Completion and Extract Messages

from ras_commander import init_ras_project, HdfResultsPlan

# Initialize project
init_ras_project("C:/Projects/MyModel", "6.6")

# Extract compute messages (handles HDF + .txt fallback automatically)
messages = HdfResultsPlan.get_compute_messages("01")

# Check if plan has results (runtime data exists only for completed plans)
runtime = HdfResultsPlan.get_runtime_data("01")
is_complete = runtime is not None

if is_complete:
    print(f"Plan completed in {runtime['Complete Process (hr)'].values[0]:.2f} hours")
else:
    print("Plan has not been executed or did not complete")

API Reference

HdfResultsPlan.get_compute_messages()

Purpose: Extract raw computation messages from HDF file

Signature:

@staticmethod
@log_call
@standardize_input(file_type='plan_hdf')
def get_compute_messages(hdf_path: Path) -> str

Parameters:

• hdf_path

  : Plan HDF file path OR plan number string (e.g., "01")

Returns: String containing all computation messages, empty string if unavailable

Fallback Behavior:

1. First attempts: HDF path

   /Results/Summary/Compute Messages (text)

2. Fallback:

   .txt

   file via RasControl (for pre-6.x HEC-RAS)

Example:

messages = HdfResultsPlan.get_compute_messages("01")
print(len(messages))  # Character count

HdfResultsPlan.get_compute_messages_hdf_only()

Purpose: Extract compute messages WITHOUT RasControl/COM fallback

Use When: Automated workflows where COM locking is problematic

Fallback Order:

1. HDF

   /Results/Summary/Compute Messages (text)

2. {plan_file}.computeMsgs.txt

   (HEC-RAS 6.x+)

3. {plan_file}.comp_msgs.txt

   (HEC-RAS 5.x)

HdfResultsPlan.get_runtime_data()

Purpose: Extract detailed performance metrics for completed plans

Returns: DataFrame with columns:

• Plan Name

  ,

  File Name

• Simulation Start Time

  ,

  Simulation End Time

• Simulation Duration (s)

  ,

  Simulation Time (hr)

• Completing Geometry (hr)

  ,

  Preprocessing Geometry (hr)

• Completing Event Conditions (hr)

  ,

  Unsteady Flow Computations (hr)

• Complete Process (hr)

• Unsteady Flow Speed (hr/hr)

  ,

  Complete Process Speed (hr/hr)

Returns None if: Plan not executed or did not complete

Common Error Patterns

Critical Errors (Plan Failed)

Unable to open geometry

Boundary condition not found

ERROR: HDF file cannot be opened

Unrecoverable error

Stability Warnings (Plan May Succeed)

Unsteady flow time step reduction

Exceeded maximum iterations

Time step reduced below minimum

Solution did not converge

Informational Messages

Complete Process

Writing Results

Completing Geometry

Completing Event Conditions

Message Parsing Pattern

Basic Severity Classification

def classify_message_severity(message_line: str) -> str:
    """Classify a compute message line by severity."""
    line_upper = message_line.upper()

    # CRITICAL - plan likely failed
    if any(x in line_upper for x in [
        'UNRECOVERABLE', 'FATAL', 'UNABLE TO OPEN',
        'CANNOT OPEN', 'ERROR:', 'FAILED'
    ]):
        return 'CRITICAL'

    # ERROR - significant issue
    if any(x in line_upper for x in [
        'ERROR', 'NOT FOUND', 'EXCEEDED MAXIMUM'
    ]):
        return 'ERROR'

    # WARNING - potential issue
    if any(x in line_upper for x in [
        'WARNING', 'TIME STEP REDUCTION', 'DID NOT CONVERGE',
        'INSTABILITY', 'REDUCED'
    ]):
        return 'WARNING'

    # INFO - normal operation
    return 'INFO'

Extract Structured Diagnostics

def parse_compute_messages(raw_messages: str) -> dict:
    """Parse raw compute messages into structured format."""
    lines = raw_messages.strip().split('\n')

    result = {
        'critical': [],
        'errors': [],
        'warnings': [],
        'info': [],
        'is_complete': False,
        'total_lines': len(lines)
    }

    for line in lines:
        line = line.strip()
        if not line:
            continue

        severity = classify_message_severity(line)

        if severity == 'CRITICAL':
            result['critical'].append(line)
        elif severity == 'ERROR':
            result['errors'].append(line)
        elif severity == 'WARNING':
            result['warnings'].append(line)
        else:
            result['info'].append(line)

        # Check for completion marker
        if 'COMPLETE PROCESS' in line.upper():
            result['is_complete'] = True

    return result

Output Schema for Orchestrators

When reporting compute message analysis, use this structured format:

## Compute Messages Analysis

### Execution Status
- **Plan**: {plan_number}
- **HDF File**: {hdf_filename}
- **Completed**: Yes/No
- **Duration**: X.XX hours (if completed)

### Messages by Severity

**CRITICAL ({count})**:
- {critical message 1}
- {critical message 2}

**ERRORS ({count})**:
- {error message 1}

**WARNINGS ({count})**:
- {warning message 1}

**INFO ({count})**: {count} informational messages (omitted for brevity)

### Diagnostics
- {Actionable interpretation}
- {Recommended next steps}

### Performance (if completed)
- Simulation Time: X.X hours
- Compute Time: X.X hours
- Speed Ratio: X.X hr/hr

Complete Workflow Example

from ras_commander import init_ras_project, HdfResultsPlan
from pathlib import Path

def analyze_plan_execution(project_path: str, ras_version: str, plan_number: str):
    """Comprehensive compute message analysis for a plan."""

    # Initialize
    init_ras_project(project_path, ras_version)

    # Get compute messages
    messages = HdfResultsPlan.get_compute_messages(plan_number)

    # Get runtime data (None if not complete)
    runtime = HdfResultsPlan.get_runtime_data(plan_number)

    # Parse messages
    parsed = parse_compute_messages(messages)

    # Build report
    report = {
        'plan': plan_number,
        'completed': runtime is not None,
        'critical_count': len(parsed['critical']),
        'error_count': len(parsed['errors']),
        'warning_count': len(parsed['warnings']),
        'info_count': len(parsed['info']),
        'runtime_hours': None,
        'speed_ratio': None
    }

    if runtime is not None:
        report['runtime_hours'] = runtime['Complete Process (hr)'].values[0]
        report['speed_ratio'] = runtime['Complete Process Speed (hr/hr)'].values[0]

    # Print summary
    print(f"Plan {plan_number}: ", end='')
    if report['completed']:
        print(f"COMPLETE in {report['runtime_hours']:.2f}h ({report['speed_ratio']:.1f}x speed)")
    else:
        print("NOT COMPLETE")

    if parsed['critical']:
        print(f"  CRITICAL: {len(parsed['critical'])} issues")
        for msg in parsed['critical'][:3]:  # Show first 3
            print(f"    - {msg[:80]}...")

    if parsed['warnings']:
        print(f"  WARNINGS: {len(parsed['warnings'])} issues")

    return report

# Usage
report = analyze_plan_execution("C:/Projects/Muncie", "6.6", "01")

Integration with Results Analyst Agent

When delegating compute message analysis to the Results Analyst Agent:

Context to Provide:

1. Plan number and HDF path
2. Expected completion status
3. Any known issues or concerns

Expected Output:

1. Structured diagnostics following output schema
2. Severity classification of all messages
3. Actionable recommendations

Example Delegation:

Task(
    subagent_type="results-analyst",
    model="sonnet",
    prompt="""
    Analyze compute messages for plan 01.

    Context files:
    - agent_tasks/.agent/STATE.md

    Task: Extract compute messages from examples/Muncie.p01.hdf,
    classify by severity, and provide diagnostics.

    Write findings to: .claude/outputs/results-analyst/compute-analysis.md
    """
)

Common Issues

Empty Messages

Cause: Plan not executed yet, or HDF missing Results/Summary group

Solution: Check if plan has been executed:

runtime = HdfResultsPlan.get_runtime_data("01")
if runtime is None:
    print("Plan has not been executed - execute with RasCmdr.compute_plan('01')")

COM Locking in Automated Workflows

Cause:

get_compute_messages()

Solution: Use

get_compute_messages_hdf_only()

# Safe for parallel/automated workflows
messages = HdfResultsPlan.get_compute_messages_hdf_only("01")

Partial Results

Cause: Plan crashed mid-execution

Symptoms:

• get_runtime_data()

  returns None

• get_compute_messages()

  shows partial output
• HDF file exists but is incomplete

Diagnosis: Check for error patterns at end of messages

Primary Sources

For complete details, navigate to:

1. ras_commander/hdf/HdfResultsPlan.py

   (lines 717-914)

   • get_compute_messages()

     implementation

   • get_compute_messages_hdf_only()

     implementation
   • HDF path

     /Results/Summary/Compute Messages (text)

2. ras_commander/hdf/HdfResultsPlan.py

   (lines 200-305)

   • get_runtime_data()

     implementation
   • Process time extraction
   • Speed calculations

3. examples/400_1d_hdf_data_extraction.ipynb

   • Working compute message extraction example
   • Output formatting patterns

4. ras_commander/hdf/AGENTS.md

   • HDF class organization
   • Decorator patterns
   • File type expectations

See Also

• hecras_extract_results: Full HDF results extraction patterns
• hecras_compute_plans: Plan execution that generates compute messages
• qa_repair_geometry: Fixing geometry errors found in messages