donpat1to/WebScraper
1
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects 1 Releases Wiki Activity

added integrity check to cleanse functions

Browse Source
This commit is contained in:
Patrick Mahnke-Hartmann
2026-01-10 18:42:39 +01:00
parent 766eb803f1
commit ac1345798d
4 changed files with 74 additions and 153 deletions
Download Patch File Download Diff File Expand all files Collapse all files

152
src/corporate/update_companies_cleanse.rs
View File

@@ -3,6 +3,7 @@ use super::{helpers::*, types::*};
use crate::config::Config;
use crate::corporate::checkpoint_helpers;
use crate::util::directories::DataPaths;
use crate::util::integrity::{DataStage, StateManager, file_reference};
use crate::util::logger;
use crate::scraper::yahoo::{YahooClientPool, QuoteSummaryModule};
@@ -14,7 +15,6 @@ use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
use tokio::fs::{File, OpenOptions};
use tokio::io::{AsyncBufReadExt, AsyncWriteExt, BufReader};
use futures::stream::{FuturesUnordered, StreamExt};
use serde_json::json;
use tokio::sync::mpsc;
/// Result of processing a single company
@@ -33,11 +33,6 @@ enum LogCommand {
Shutdown,
}
/// Result from processing a single company with priority
struct CompanyTaskResult {
company: CompanyCrossPlatformInfo,
result: CompanyProcessResult,
}
/// Cleansing function to remove companies with missing essential yahoo data for integrity
pub async fn companies_yahoo_cleansed_no_data(paths: &DataPaths) -> Result<usize, anyhow::Error> {
@@ -51,36 +46,23 @@ pub async fn companies_yahoo_cleansed_no_data(paths: &DataPaths) -> Result<usize
logger::log_warn("companies.jsonl not found, skipping cleansing").await;
return Ok(0);
}
let manager = StateManager::new(&state_path, &data_path.to_path_buf());
let step_name = "yahoo_companies_cleansed_no_data";
let content_reference = file_reference(&output_path);
if state_path.exists() {
let state_content = tokio::fs::read_to_string(&state_path).await?;
if manager.is_step_valid(step_name).await? {
let output_content = tokio::fs::read_to_string(&output_path).await?;
let count = output_content.lines()
.filter(|line| !line.trim().is_empty())
.count();
for line in state_content.lines() {
if line.trim().is_empty() {
continue;
}
if let Ok(state) = serde_json::from_str::<serde_json::Value>(line) {
if state.get("yahoo_companies_cleansed_no_data").and_then(|v| v.as_bool()).unwrap_or(false) {
logger::log_info(" Yahoo companies cleansing already completed, reading existing file...").await;
if output_path.exists() {
let output_content = tokio::fs::read_to_string(&output_path).await?;
let count = output_content.lines()
.filter(|line| !line.trim().is_empty())
.count();
logger::log_info(&format!(" ✓ Found {} companies in companies_yahoo.jsonl", count)).await;
return Ok(count);
} else {
logger::log_warn(" State indicates completion but companies_yahoo.jsonl not found, re-running...").await;
break;
}
}
}
}
logger::log_info(&format!(" ✓ Found {} companies in companies_yahoo.jsonl", count)).await;
return Ok(count);
}
logger::log_info(" Cleansing companies with missing Yahoo data...").await;
logger::log_info(&format!(" Reading from: {:?}", input_path)).await;
logger::log_info(&format!(" Writing to: {:?}", output_path)).await;
@@ -141,18 +123,17 @@ pub async fn companies_yahoo_cleansed_no_data(paths: &DataPaths) -> Result<usize
total_count, valid_count, removed_count
)).await;
let yahoo_companies = json!({
"yahoo_companies_cleansed_no_data": true,
"completed_at": chrono::Utc::now().to_rfc3339(),
});
let mut state_file = File::create(&state_path).await?;
let state_line = serde_json::to_string(&yahoo_companies)?;
state_file.write_all(state_line.as_bytes()).await?;
state_file.write_all(b"\n").await?;
state_file.flush().await?;
logger::log_info(&format!(" ✓ State file created at: {:?}", state_path)).await;
// Track completion with:
// - Content reference: All event directories
// - Data stage: Data (7-day TTL by default)
// - Dependencies: Depends on cleaned companies data
manager.update_entry(
step_name.to_string(),
content_reference,
DataStage::Data,
vec!["yahoo_companies_cleansed".to_string()], // Dependency
None, // Use default TTL (7 days for Data stage)
).await?;
Ok(valid_count)
}
@@ -199,6 +180,22 @@ pub async fn companies_yahoo_cleansed_low_profile(
return Ok(0);
}
let manager = StateManager::new(&state_path, &data_path.to_path_buf());
let step_name = "yahoo_companies_cleansed_no_data";
let content_reference = file_reference(&checkpoint_path);
if manager.is_step_valid(step_name).await? {
let checkpoint_content = tokio::fs::read_to_string(&checkpoint_path).await?;
let count = checkpoint_content.lines()
.filter(|line| !line.trim().is_empty())
.count();
logger::log_info(&format!(" ✓ Found {} companies in companies_yahoo_cleaned.jsonl", count)).await;
return Ok(count);
}
logger::log_info(" Cleansing companies with low Yahoo profile...").await;
if state_path.exists() {
let state_content = tokio::fs::read_to_string(&state_path).await?;
@@ -505,7 +502,7 @@ pub async fn companies_yahoo_cleansed_low_profile(
}
match task_result {
Ok(Ok(Some(_result))) => {
Ok(Ok(_)) => {
// Success - spawn next task
if let Some(company) = pending.pop() {
spawn_validation_task(
@@ -524,25 +521,6 @@ pub async fn companies_yahoo_cleansed_low_profile(
);
}
}
Ok(Ok(None)) => {
// Filtered or failed - spawn next task
if let Some(company) = pending.pop() {
spawn_validation_task(
company,
&yahoo_pool,
&paths,
&write_tx,
shutdown_flag,
&processed,
&valid_count,
&filtered_low_cap,
&filtered_no_price,
&failed_count,
total,
&mut tasks,
);
}
}
Ok(Err(e)) => {
// Processing error
logger::log_error(&format!("Company processing error: {}", e)).await;
@@ -646,24 +624,17 @@ pub async fn companies_yahoo_cleansed_low_profile(
// Shutdown Yahoo pool
yahoo_pool.shutdown().await?;
// Write completion milestone to state.jsonl
let state_path = data_path.join("state.jsonl");
let yahoo_low_profile = json!({
"yahoo_companies_cleansed_low_profile": true,
"completed_at": chrono::Utc::now().to_rfc3339(),
});
let mut state_file = OpenOptions::new()
.create(true)
.append(true)
.open(&state_path)
.await?;
let state_line = serde_json::to_string(&yahoo_low_profile)?;
state_file.write_all(state_line.as_bytes()).await?;
state_file.write_all(b"\n").await?;
state_file.flush().await?;
logger::log_info(&format!(" ✓ State milestone saved to: {:?}", state_path)).await;
// Track completion with:
// - Content reference: All event directories
// - Data stage: Data (7-day TTL by default)
// - Dependencies: Depends on cleaned companies data
manager.update_entry(
step_name.to_string(),
content_reference,
DataStage::Data,
vec!["yahoo_companies_cleansed".to_string()], // Dependency
None, // Use default TTL (7 days for Data stage)
).await?;
Ok(final_count)
}
@@ -681,7 +652,7 @@ fn spawn_validation_task(
filtered_no_price: &Arc<AtomicUsize>,
failed_count: &Arc<AtomicUsize>,
total: usize,
tasks: &mut FuturesUnordered<tokio::task::JoinHandle<anyhow::Result<Option<CompanyTaskResult>>>>,
tasks: &mut FuturesUnordered<tokio::task::JoinHandle<anyhow::Result<Option<()>>>>,
) {
let yahoo_pool_clone = Arc::clone(yahoo_pool);
let paths_clone = Arc::clone(paths);
@@ -705,36 +676,29 @@ fn spawn_validation_task(
&*paths_clone,
).await;
let task_result = match result {
match result {
CompanyProcessResult::Valid(validated_company) => {
// Send to writer
let _ = write_tx_clone.send(LogCommand::Write(validated_company.clone())).await;
let _ = write_tx_clone.send(LogCommand::Write(validated_company)).await;
valid_count_clone.fetch_add(1, Ordering::SeqCst);
Some(CompanyTaskResult {
company: validated_company.clone(),
result: CompanyProcessResult::Valid(validated_company),
})
}
CompanyProcessResult::FilteredLowCap { name, market_cap } => {
filtered_low_cap_clone.fetch_add(1, Ordering::SeqCst);
if filtered_low_cap_clone.load(Ordering::SeqCst) <= 10 {
logger::log_info(&format!(" Filtered {} - low market cap: {:.0} EUR", name, market_cap)).await;
}
None
}
CompanyProcessResult::FilteredNoPrice { name } => {
filtered_no_price_clone.fetch_add(1, Ordering::SeqCst);
if filtered_no_price_clone.load(Ordering::SeqCst) <= 10 {
logger::log_info(&format!(" Filtered {} - no recent price data", name)).await;
}
None
}
CompanyProcessResult::Failed { company: failed_company, error, is_transient: _ } => {
failed_count_clone.fetch_add(1, Ordering::SeqCst);
logger::log_warn(&format!(" Failed to process '{}': {}", failed_company.name, error)).await;
None
}
};
}
// Progress reporting
let current = processed_clone.fetch_add(1, Ordering::SeqCst) + 1;
@@ -749,7 +713,7 @@ fn spawn_validation_task(
)).await;
}
Ok(task_result)
Ok(None::<()>)
});
tasks.push(task);