Optimize CSV validation and header mapping logic if header is enabled

Makefile

@@ -93,7 +93,11 @@ BENCH_ROWS_SRC    ?= 2000
 BENCH_CSV_PATH    := ./build/bench/$(BENCH_COLS)_$(BENCH_ROWS_SRC)_000.csv
 BENCH_CSV         := --csv='$(BENCH_CSV_PATH)'
 BENCH_FLAGS       := --debug --profile --report=text -vvv
-BENCH_SCHEMAS   := --schema='./tests/Benchmarks/bench_*.yml'
+BENCH_SCHEMAS_ALL := --schema='./tests/Benchmarks/bench_*.yml'
+BENCH_SCHEMAS_0   := --schema='./tests/Benchmarks/bench_0_*.yml'
+BENCH_SCHEMAS_1   := --schema='./tests/Benchmarks/bench_1_*.yml'
+BENCH_SCHEMAS_2   := --schema='./tests/Benchmarks/bench_2_*.yml'
+BENCH_SCHEMAS_3   := --schema='./tests/Benchmarks/bench_3_*.yml'
 
 
 bench: ##@Benchmarks Run all benchmarks
@@ -109,23 +113,37 @@ bench-create-csv: ##@Benchmarks Create CSV file
 
 
 bench-docker: ##@Benchmarks Run CSV file with Docker
+	@docker run --rm  $(DOCKER_IMAGE) --ansi --version
 	@echo "::group::Quickest"
-	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) --schema='./tests/Benchmarks/bench_0_*.yml' $(BENCH_FLAGS)
+	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) $(BENCH_SCHEMAS_0) $(BENCH_FLAGS)
 	@echo "::endgroup::"
 	@echo "::group::Minimum"
-	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) --schema='./tests/Benchmarks/bench_1_*.yml' $(BENCH_FLAGS)
+	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) $(BENCH_SCHEMAS_1) $(BENCH_FLAGS)
 	@echo "::endgroup::"
 	@echo "::group::Realistic"
-	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) --schema='./tests/Benchmarks/bench_2_*.yml' $(BENCH_FLAGS)
+	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) $(BENCH_SCHEMAS_2) $(BENCH_FLAGS)
 	@echo "::endgroup::"
 	@echo "::group::All aggregations at once"
-	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) --schema='./tests/Benchmarks/bench_3_*.yml' $(BENCH_FLAGS)
+	-$(BLUEPRINT_DOCKER) $(BENCH_CSV) $(BENCH_SCHEMAS_3) $(BENCH_FLAGS)
 	@echo "::endgroup::"
 
 
 bench-phar: ##@Benchmarks Run CSV file with Phar
-	-$(BLUEPRINT_PHAR) $(BENCH_CSV) $(BENCH_SCHEMAS) $(BENCH_FLAGS)
+	./build/csv-blueprint.phar --ansi --version
+	@echo "::group::Quickest"
+	-$(BLUEPRINT_PHAR) $(BENCH_CSV) $(BENCH_SCHEMAS_0) $(BENCH_FLAGS)
+	@echo "::endgroup::"
+	@echo "::group::Minimum"
+	-$(BLUEPRINT_PHAR) $(BENCH_CSV) $(BENCH_SCHEMAS_1) $(BENCH_FLAGS)
+	@echo "::endgroup::"
+	@echo "::group::Realistic"
+	-$(BLUEPRINT_PHAR) $(BENCH_CSV) $(BENCH_SCHEMAS_2) $(BENCH_FLAGS)
+	@echo "::endgroup::"
+	@echo "::group::All aggregations at once"
+	-$(BLUEPRINT_PHAR) $(BENCH_CSV) $(BENCH_SCHEMAS_3) $(BENCH_FLAGS)
+	@echo "::endgroup::"
 
 
 bench-php: ##@Benchmarks Run CSV file with classic PHP binary
+	$(PHP_BIN) ./csv-blueprint --ansi --version
 	-$(BLUEPRINT) $(BENCH_CSV) $(BENCH_SCHEMAS) $(BENCH_FLAGS)

src/Csv/Column.php

@@ -107,6 +107,11 @@ public function validateCell(string $cellValue, int $line = Error::UNDEFINED_LIN
         return $this->getValidator()->validateCell($cellValue, $line);
     }
 
+    public function setId(int $realIndex): void
+    {
+        $this->id = $realIndex;
+    }
+
     private function prepareRuleSet(string $schemaKey): array
     {
         $rules = [];

src/Csv/CsvFile.php

@@ -65,7 +65,9 @@ public function getHeader(): array
             if ($this->structure->isHeader() && !$this->isEmpty) {
                 // TODO: add handler for empty file
                 // League\Csv\SyntaxError : The header record does not exist or is empty at offset: `0
-                $this->header = $this->reader->getHeader();
+                $this->header = $this->getRecordsChunk(0, 1)->first();
+            } else {
+                $this->header = \range(0, \count($this->getRecordsChunk(0, 1)->first()) - 1);
             }
         }
 
@@ -74,12 +76,12 @@ public function getHeader(): array
 
     public function getRecords(): \Iterator
     {
-        return $this->reader->getRecords($this->getHeader());
+        return $this->reader->getRecords([]);
     }
 
     public function getRecordsChunk(int $offset = 0, int $limit = -1): TabularDataReader
     {
-        return Statement::create(null, $offset, $limit)->process($this->reader, $this->getHeader());
+        return Statement::create(null, $offset, $limit)->process($this->reader, []); // No headers is required!
     }
 
     public function validate(bool $quickStop = false): ErrorSuite
@@ -92,13 +94,39 @@ public function getRealColumNumber(): int
         return \count($this->getRecordsChunk(0, 1)->first());
     }
 
+    public function getSchema(): Schema
+    {
+        return $this->schema;
+    }
+
+    /**
+     * @return Column[]
+     */
+    public function getColumnsMappedByHeader(): array
+    {
+        $map = [];
+
+        $realHeader = $this->getHeader();
+        foreach ($realHeader as $realIndex => $realColumn) {
+            $realIndex = (int)$realIndex;
+            $schemaColumn = $this->schema->getColumn($realColumn);
+
+            if ($schemaColumn !== null) {
+                $schemaColumn->setId($realIndex);
+                $map[$realIndex] = $schemaColumn;
+            }
+        }
+
+        return $map;
+    }
+
     private function prepareReader(): LeagueReader
     {
         $reader = LeagueReader::createFromPath($this->csvFilename)
             ->setDelimiter($this->structure->getDelimiter())
             ->setEnclosure($this->structure->getEnclosure())
             ->setEscape($this->structure->getQuoteChar())
-            ->setHeaderOffset($this->structure->isHeader() ? 0 : null);
+            ->setHeaderOffset(null); // It's important to set it to null to optimize memory usage!
 
         if ($this->structure->isBom()) {
             $reader->includeInputBOM();

src/Schema.php

@@ -86,25 +86,6 @@ public function getColumns(): array
         return $this->columns;
     }
 
-    /**
-     * @return Column[]|null[]
-     * @phan-suppress PhanPartialTypeMismatchReturn
-     */
-    public function getColumnsMappedByHeader(array $header): array
-    {
-        $map = [];
-
-        if ($this->getCsvStructure()->isHeader()) {
-            foreach ($header as $headerName) {
-                $map[$headerName] = $this->columns[$headerName] ?? null;
-            }
-        } else {
-            return $this->getColumns();
-        }
-
-        return $map;
-    }
-
     public function getColumn(int|string $columNameOrId): ?Column
     {
         if (\is_int($columNameOrId)) {
@@ -113,10 +94,6 @@ public function getColumn(int|string $columNameOrId): ?Column
             $column = $this->getColumns()[$columNameOrId] ?? null;
         }
 
-        if ($column === null) {
-            throw new Exception("Column \"{$columNameOrId}\" not found in schema \"{$this->filename}\"");
-        }
-
         return $column;
     }
 
@@ -154,6 +131,11 @@ public function getData(): AbstractData
         return clone $this->data;
     }
 
+    public function getSchemaHeader(): array
+    {
+        return \array_keys($this->getColumns());
+    }
+
     /**
      * @return Column[]
      */

src/Utils.php

@@ -63,8 +63,10 @@ public static function debug(int|string $message): void
 
     public static function debugSpeed(string $messPrefix, int $lines, float $startTimer): void
     {
-        $kiloLines = \round(($lines / (\microtime(true) - $startTimer)) / 1000);
-        self::debug("{$messPrefix} <blue>" . \number_format($kiloLines) . 'K</blue> lines/sec');
+        if (\defined('DEBUG_MODE')) {
+            $kiloLines = \round(($lines / (\microtime(true) - $startTimer)) / 1000);
+            self::debug("{$messPrefix} <blue>" . \number_format($kiloLines) . 'K</blue> lines/sec');
+        }
     }
 
     public static function kebabToCamelCase(string $input): string

src/Validators/ValidatorCsv.php

@@ -106,15 +106,13 @@ private function validateHeader(bool $quickStop = false): ErrorSuite
     private function validateLines(bool $quickStop = false): ErrorSuite
     {
         $errors = new ErrorSuite();
-        $realColumns = $this->schema->getColumnsMappedByHeader($this->csv->getHeader());
+        $mappedColumns = $this->csv->getColumnsMappedByHeader();
+        $isHeaderEnabled = $this->schema->getCsvStructure()->isHeader();
 
-        foreach ($realColumns as $column) {
-            $columValues = [];
-            if ($column === null) {
-                continue;
-            }
+        foreach ($mappedColumns as $columnIndex => $column) {
+            $messPrefix = "<i>Column</i> \"{$column->getHumanName()}\" -"; // System message prefix. Debug only!
 
-            $messPrefix = "<i>Column</i> \"{$column->getHumanName()}\" -";
+            $columValues = [];
 
             Utils::debug("{$messPrefix} Column start");
             $colValidator = $column->getValidator();
@@ -138,30 +136,34 @@ private function validateLines(bool $quickStop = false): ErrorSuite
             $lineCounter = 0;
             $startTimer = \microtime(true);
             foreach ($this->csv->getRecords() as $line => $record) {
+                if ($isHeaderEnabled && $line === 0) {
+                    continue;
+                }
+
                 $lineCounter++;
                 $lineNum = (int)$line + 1;
 
                 if ($isRules) { // Time optimization
-                    if (!isset($record[$column->getKey()])) {
+                    if (!isset($record[$columnIndex])) {
                         $errors->addError(
                             new Error(
                                 'csv.column',
-                                "Column index:{$column->getKey()} not found",
+                                "Column index:{$columnIndex} not found",
                                 $column->getHumanName(),
                                 $lineNum,
                             ),
                         );
                     } else {
-                        $errors->addErrorSuit($colValidator->validateCell($record[$column->getKey()], $lineNum));
+                        $errors->addErrorSuit($colValidator->validateCell($record[$columnIndex], $lineNum));
                     }
 
                     if ($quickStop && $errors->count() > 0) {
                         return $errors;
                     }
                 }
 
-                if ($isAggRules && isset($record[$column->getKey()])) {  // Time & memory optimization
-                    $columValues[] = ValidatorColumn::prepareValue($record[$column->getKey()], $aggInputType);
+                if ($isAggRules && isset($record[$columnIndex])) {  // Time & memory optimization
+                    $columValues[] = ValidatorColumn::prepareValue($record[$columnIndex], $aggInputType);
                 }
             }
             Utils::debug("{$messPrefix} Lines <yellow>" . \number_format($lineCounter) . '</yellow>');
@@ -213,10 +215,9 @@ private function validateColumn(bool $quickStop): ErrorSuite
         $errors = new ErrorSuite();
 
         if ($this->schema->getCsvStructure()->isHeader()) {
-            $realColumns = $this->schema->getColumnsMappedByHeader($this->csv->getHeader());
-            $schemaColumns = $this->schema->getColumns();
-
-            $notFoundColums = \array_diff(\array_keys($schemaColumns), \array_keys($realColumns));
+            $realColumns = $this->csv->getHeader();
+            $schemaColumns = $this->schema->getSchemaHeader();
+            $notFoundColums = \array_diff($schemaColumns, $realColumns);
 
             if (\count($notFoundColums) > 0) {
                 $error = new Error(

tests/Benchmarks/bench_0_quickest.yml

@@ -12,9 +12,7 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - rules:
+  - name: id
+    rules:
       not_empty: true

tests/Benchmarks/bench_0_quickest_agg.yml

@@ -12,9 +12,7 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - aggregate_rules:
+  - name: id
+    aggregate_rules:
       count: 0

tests/Benchmarks/bench_0_quickest_combo.yml

@@ -12,11 +12,9 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - rules:
+  - name: id
+    rules:
       not_empty: true
     aggregate_rules:
       count: 0

tests/Benchmarks/bench_1_mini.yml

@@ -12,10 +12,8 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - rules:
+  - name: id
+    rules:
       not_empty: true
       is_int: true

tests/Benchmarks/bench_1_mini_agg.yml

@@ -12,10 +12,8 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - aggregate_rules:
+  - name: id
+    aggregate_rules:
       average: 0
       count: 0

tests/Benchmarks/bench_1_mini_combo.yml

@@ -12,11 +12,9 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - rules:
+  - name: id
+    rules:
       not_empty: true
       is_int: true
     aggregate_rules:

tests/Benchmarks/bench_2_realistic.yml

@@ -12,11 +12,9 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - rules:
+  - name: id
+    rules:
       not_empty: true
       length_max: 100
       is_int: true

tests/Benchmarks/bench_2_realistic_agg.yml

@@ -12,11 +12,9 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - aggregate_rules:
+  - name: id
+    aggregate_rules:
       is_unique: true
       sorted: [ desc, natural ]
       count: 0

tests/Benchmarks/bench_2_realistic_combo.yml

@@ -12,11 +12,9 @@
 
 filename_pattern: /.csv$/i
 
-csv:
-  header: false
-
 columns:
-  - rules:
+  - name: id
+    rules:
       not_empty: true
       length_max: 100
       is_int: true