smarter_csv 1.17.3 → 1.18.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ec50e8539c6872f9c86c25eabc2982e39846ad07dc5a21021fc687c7661f8084
-  data.tar.gz: 977ce04d8dd225b6042ea03ad0c174305f3ea122340fad052e2c2ada440d6400
+  metadata.gz: 3335e39a1c0792f01df9e95401c7f3885c49a0d64eeb9c76e5c20e25d01a62f5
+  data.tar.gz: e43f00228777b56fc1ee0814a74acaa6a23c51fe8da6f64e42ad92fe1b54002f
 SHA512:
-  metadata.gz: 0452dc7f15ab31b0cdfad83ca718e17e6456cf6c9826d177e606c5924f3ec72a155c86ee6f9f938540fe3b2ed8f694a981c95cf775b5a38d7f7e44318bc453a3
-  data.tar.gz: c1c9732d6d4393fb2ffa995f0c7bb73cd60f566132d487a86007f8a5a623257365c2324ed894a5b38d681df7cfab67069d9fa2e61fd525ba51675954ddadad7a
+  metadata.gz: 2abcd136f30d284c3c27cbd2b6c9782aec4235ec62cc27ea7620380ae9efc889f9f1c05c10ad957d6e7c84d65be75d35d96d4ac59ba5780dc5e65e0151c661e6
+  data.tar.gz: 6c61062c08d0a89dea2c91a7faafd88f6d09c88015ad8c7f4facb2eb474b44b18e0c5eaffc36fa5e2885f0b2c02487497f2460d66ce2433f62c09bf42d92a4ce

data/CHANGELOG.md

@@ -2,7 +2,56 @@
 # SmarterCSV 1.x Change Log
 
 > [!TIP]
-> **Upgrading?** The [SmarterCSV Upgrade Wizard](https://tilo.github.io/smarter_csv/upgrade_wizard.html) walks you through what (if anything) you need to change for your specific version. Most hops do not require any changes.
+> **Upgrading?** The [SmarterCSV Upgrade Wizard](https://tilo.github.io/smarter_csv/upgrade_wizard.html) walks you through what (if anything) you need to change for your specific version. Most steps do not require any changes.
+
+## 1.18.0 (2026-06-17)
+
+This release is focused on both performance and the introduction of automatic conversion of decimals to big_decimal or float, preserving the precision, and also supporting scientific notation.
+
+⚠️ This version is particularly interesting if you have geolocation, scientific, or high-precision data.
+
+### New Features
+
+  - **`decimal_precision` option** (`:auto` default, or `:float` / `:bigdecimal`) — controls how decimal values are converted. `:auto` returns a `Float` unless the value carries more than 16 significant digits, in which case it returns a `BigDecimal` so no precision is lost; `:float` always returns `Float`; `:bigdecimal` always returns `BigDecimal`. Integers are unaffected (always `Integer`). Works identically on the C and Ruby paths. (Ruby's standard-library CSV has no high-precision option — its `:numeric`/`:float` converters use `Float()` and lose precision.)
+  - **Float** conversion on the C path now uses the fast **Eisel-Lemire** algorithm (fast_float, vendored) for mantissas up to 19 significant digits — correctly rounded, bit-for-bit identical to `String#to_f` — with a `strtod` fallback beyond that (more than 19 digits / extreme exponents). High-precision values that become `BigDecimal` under `:auto`/`:bigdecimal` are parsed by Ruby's `BigDecimal`.
+
+### Behavior Changes
+
+  - **Scientific notation now converts to a number** (e.g. `"1e3"`, `"1.5e-5"`, `"6.022e23"`). Previously the Ruby path left these as Strings and the C path was inconsistent.
+  - **The C and Ruby numeric-conversion paths are now aligned.** Bare-dot forms like `".5"` and `"3."` stay Strings on **both** paths (the shared grammar requires an integer part and, when a dot is present, a fraction digit). Previously the C path converted these and the Ruby path did not.
+  - With the default `decimal_precision: :auto`, decimal values carrying more than 16 significant digits are now returned as `BigDecimal` instead of `Float`. Pass `decimal_precision: :float` to keep the previous always-`Float` behavior.
+  - `bigdecimal` is now a runtime dependency (it is no longer a default gem on Ruby 3.4+).
+
+### Performance
+
+  The C-accelerated path is faster across the board, **up to ~1.5× on the right shapes** — numeric-heavy data and backslash-escaped quoted fields — and ~1.04–1.08× on typical files.
+
+  - Eisel-Lemire (Mushtak-Lemire) algorithm on the C path to convert decimals to `Float` or `BigDecimal`. Numeric-heavy data (many float/decimal columns) parses significantly faster.
+  - SIMD scanner for backslash-escaped quoted fields (C-path), using NEON (arm64) and SSE2 (x86-64) with a scalar fallback. Speeds up `quote_escaping: :backslash` parsing of long quoted fields.
+
+  | File                            | C-path                           | driver                |
+  |---------------------------------|----------------------------------|-----------------------|
+  | backslash_long_fields_60k       | 1.48× faster (0.1880s → 0.1273s) | SIMD quote/backslash scanner |
+  | sensor_data_50krows_50cols      | 1.40× faster (0.2763s → 0.1975s) | Eisel-Lemire numeric conversion |
+
+### Improvements
+
+  - Improved robustness of symbol-valued enum option processing.
+
+### Tests
+
+  - added parity tests for long quoted-field scanning across 16-byte boundaries, running on both the C and Ruby paths.
+  - added tests for string-to-symbol coercion of the enum options.
+
+
+
+## 1.17.4 (2026-06-03)
+
+### Bug Fix
+
+  - fixed [Issue #337](https://github.com/tilo/smarter_csv/issues/337): `Pathname` input no longer worked (regression since 1.17.0); passing a `Pathname` raised `NoMethodError: private method 'gets' called`. `SmarterCSV` now opens any path-like input (`String` or `Pathname`) and reads directly from any already-open IO. Thanks to [Alex Shenia](https://github.com/alexshenia)
+
+
 
 ## 1.17.3 (2026-05-26)
 

data/CONTRIBUTORS.md

@@ -1,4 +1,4 @@
-# A Big Thank You to all 64 Contributors!!
+# A Big Thank You to all 65 Contributors!!
 
 
 A Big Thank you to everyone who filed issues, sent comments, and who contributed with pull requests:
@@ -67,3 +67,4 @@ A Big Thank you to everyone who filed issues, sent comments, and who contributed
  * [Paho Lurie-Gregg](https://github.com/paholg)
  * [Jonas Staškevičius](https://github.com/pirminis)
  * [conorg](https://github.com/conorg)
+ * [Alex Shenia](https://github.com/alexshenia)

data/README.md

@@ -4,7 +4,7 @@
 ![Gem Version](https://img.shields.io/gem/v/smarter_csv) [![codecov](https://codecov.io/gh/tilo/smarter_csv/branch/main/graph/badge.svg?token=1L7OD80182)](https://codecov.io/gh/tilo/smarter_csv) [![Downloads](https://img.shields.io/gem/dt/smarter_csv)](https://rubygems.org/gems/smarter_csv) [![RubyGems](https://img.shields.io/badge/RubyGems-smarter__csv-brightgreen?logo=rubygems&logoColor=white)](https://rubygems.org/gems/smarter_csv) [![Ruby Toolbox](https://img.shields.io/badge/Ruby%20Toolbox-smarter__csv-brightgreen)](https://www.ruby-toolbox.com/projects/smarter_csv) [![Upgrade Wizard](https://img.shields.io/badge/Upgrade%20Wizard-Try%20it-2c7a2c?style=flat)](https://tilo.github.io/smarter_csv/upgrade_wizard.html)
 
 > [!TIP]
-> **Upgrading from an older version?** Use the [SmarterCSV Upgrade Wizard](https://tilo.github.io/smarter_csv/upgrade_wizard.html) to walk through what (if anything) you need to change for your specific version. Most hops do not require any changes.
+> **Upgrading from an older version?** Use the [SmarterCSV Upgrade Wizard](https://tilo.github.io/smarter_csv/upgrade_wizard.html) to walk through what (if anything) you need to change for your specific version. Most steps do not require any changes.
         
   SmarterCSV is a high-performance CSV ingestion and generation for Ruby, focused on fast end-to-end CSV ingestion of real-world data — no silent failures, no surprises, not just tokenization.
 
@@ -15,6 +15,9 @@
 
   > See [**Ruby CSV Pitfalls**](docs/ruby_csv_pitfalls.md) for 10 ways `CSV.read` silently corrupts or loses data, and how SmarterCSV handles them.
 
+  > [!TIP]
+  > **No silent precision loss (new in 1.18.0).** For scientific data, GPS/geo coordinates, and financial figures — which routinely carry 16+ significant digits — Ruby's standard CSV converts with `Float()`, so a value like `1234567890.123456789` is silently rounded to `1234567890.1234567`. SmarterCSV's default `decimal_precision: :auto` returns a `BigDecimal` for values beyond 16 significant digits (and `Float` otherwise) — full precision, no data loss. Floats are decoded with the Eisel-Lemire algorithm: correctly rounded, bit-for-bit identical to `String#to_f`.
+
   Beyond raw speed, SmarterCSV is designed to provide a significantly more convenient and developer-friendly interface than traditional CSV libraries. Instead of returning raw arrays that require substantial post-processing, SmarterCSV produces Rails-ready hashes for each row, making the data immediately usable with ActiveRecord, Sidekiq pipelines, parallel processing, and JSON-based workflows such as S3.
 
   In a Rails app, warnings auto-route through `Rails.logger` and instrumentation hooks compose with `ActiveSupport::Notifications` — no setup required. Outside Rails, warnings fall back to `$stderr` and the same APIs work without any framework dependency.
@@ -89,6 +92,8 @@ rows = SmarterCSV.process('data.csv')
 data = SmarterCSV.parse(csv_string)
 ```
 
+Numeric conversion is also more accurate: where Ruby's `:numeric`/`:float` converters round high-precision decimals through `Float()`, SmarterCSV's default `decimal_precision: :auto` returns a `BigDecimal` past 16 significant digits, so no precision is lost (pass `decimal_precision: :float` for like-for-like `Float` output).
+
 * See [**Migrating from Ruby CSV**](docs/migrating_from_csv.md) for a full comparison of options, behavior differences, and a quick-reference table.
 
 ## Examples
@@ -336,7 +341,7 @@ For reporting issues, please:
   * open a pull-request adding a test that demonstrates the issue
   * mention your version of SmarterCSV, Ruby, Rails
 
-# [A Special Thanks to all 64 Contributors!](CONTRIBUTORS.md) 🎉🎉🎉
+# [A Special Thanks to all 65 Contributors!](CONTRIBUTORS.md) 🎉🎉🎉
 
 
 ## Contributing

data/docs/data_transformations.md

@@ -156,6 +156,39 @@ data = SmarterCSV.process(file,
   convert_values_to_numeric: { only: [:quantity, :price] })
 ```
 
+Scientific notation (e.g. `"1.5e3"`, `"6.022e23"`) is recognized and converted too. Bare-dot forms like `".5"` and `"3."` are left as Strings (they are not valid numbers here). Integers and floats convert identically on the C-accelerated and pure-Ruby paths.
+
+---
+
+## `decimal_precision`
+
+**Default: `:auto`**
+
+Controls how decimal values (those with a `.` or an exponent) are converted. Integers are unaffected — they are always returned as `Integer`.
+
+| Value         | Result                                                                                  |
+|---------------|-----------------------------------------------------------------------------------------|
+| `:auto`       | `Float`, unless the value carries more than 16 significant digits — then `BigDecimal`.   |
+| `:float`      | Always `Float` (correctly rounded; matches `String#to_f`).                               |
+| `:bigdecimal` | Always `BigDecimal` (full precision).                                                    |
+
+```ruby
+# :auto (default) — keeps full precision only when needed
+SmarterCSV.process(file)
+# "3.14"                 => 3.14                              (Float)
+# "1234567890.123456789" => 0.1234567890123456789e10          (BigDecimal — >16 sig digits)
+
+# :float — always Float (faster, may lose precision on long decimals)
+SmarterCSV.process(file, decimal_precision: :float)
+# "1234567890.123456789" => 1234567890.1234567               (Float)
+
+# :bigdecimal — always BigDecimal
+SmarterCSV.process(file, decimal_precision: :bigdecimal)
+# "3.14" => 0.314e1 (BigDecimal)
+```
+
+Unlike Ruby's standard-library CSV — whose `:numeric`/`:float` converters use `Float()` and silently lose precision — `:auto` preserves high-precision decimals as `BigDecimal`. Decimal values are decoded on the C path with the Eisel-Lemire algorithm (correctly rounded, identical to `String#to_f`).
+
 ---
 
 ## `remove_empty_hashes`

data/docs/migrating_from_csv.md

@@ -223,6 +223,24 @@ rows = SmarterCSV.process('sample.csv',
   convert_values_to_numeric: { except: [:zip_code, :phone, :account_number] })
 ```
 
+**High-precision decimals — scientific data and geo coordinates.** GPS/geo coordinates, scientific measurements, and financial figures routinely carry 16+ significant digits, where Ruby's `Float()`-based conversion (`converters: :numeric` / `:float`) silently rounds the value. SmarterCSV's default `decimal_precision: :auto` returns a `BigDecimal` once a value exceeds 16 significant digits (and a `Float` otherwise), so the full value is preserved; scientific notation (`6.022e23`, `1.6e-19`) is recognized as numeric too.
+
+**With Ruby CSV (precision lost):**
+```ruby
+CSV.read('locations.csv', headers: true, converters: :float).first['lat']
+# => -122.42200352825247   ← Float() dropped the last digits of -122.422003528252475
+```
+
+**With SmarterCSV (full precision kept):**
+```ruby
+SmarterCSV.process('locations.csv').first[:lat]
+# => -0.122422003528252475e3   (BigDecimal — all 18 significant digits preserved)
+
+# Force Float everywhere, like-for-like with Ruby CSV:
+SmarterCSV.process('locations.csv', decimal_precision: :float).first[:lat]
+# => -122.42200352825247   (Float)
+```
+
 ### 3. Empty values are removed by default
 
 SmarterCSV drops key/value pairs where the value is `nil` or blank

data/docs/options.md

@@ -121,7 +121,8 @@ See [Parsing Strategy](./parsing_strategy.md) for full details on quote handling
 | Option | Default | Explanation |
 |--------|---------|-------------|
 | `:strip_whitespace` | `true` | Remove whitespace before/after values and headers. |
-| `:convert_values_to_numeric` | `true` | Convert strings containing integers or floats to the appropriate numeric type. Accepts `{except: [:key1, :key2]}` or `{only: :key3}` to limit which columns. |
+| `:convert_values_to_numeric` | `true` | Convert strings containing integers or floats (including scientific notation like `1.5e3`) to the appropriate numeric type. Accepts `{except: [:key1, :key2]}` or `{only: :key3}` to limit which columns. |
+| `:decimal_precision` | `:auto` | How decimals are converted: `:auto` returns `Float` but `BigDecimal` above 16 significant digits (no precision loss); `:float` always returns `Float`; `:bigdecimal` always returns `BigDecimal`. Integers are unaffected. |
 | `:value_converters` | `nil` | Hash of `:header => converter`; converter can be a lambda/Proc or a class implementing `self.convert(value)`. See [Value Converters](./value_converters.md). |
 | `:remove_empty_values` | `true` | Remove key/value pairs where the value is `nil`, empty, or whitespace-only — any Unicode whitespace, same as Ruby's `String#blank?`. |
 | `:remove_zero_values` | `false` | Remove key/value pairs whose value is zero — numeric `0` / `0.0`, or any textual form of zero (`"0"`, `"0.0"`, `"00.00"`, `"+0"`, `"-0.0"`, …). |

data/docs/upgrade_wizard.html

@@ -153,7 +153,7 @@ button.primary:disabled:hover { background: #e6e6e6; }
 <body>
 
 <h1>SmarterCSV Upgrade Wizard</h1>
-<p class="muted">This wizard walks you from your current version to the latest, one hop at a time.<br><br>Only the questions where you answer "Yes" will show migration steps.<br>Question answered with "No" represent risk-free upgrades.</p><br><br>
+<p class="muted">This wizard walks you from your current version to the latest, one step at a time.<br><br>Only the questions where you answer "Yes" will show migration steps.<br>Question answered with "No" represent risk-free upgrades.</p><br><br>
 
 <div id="app"></div>
 
@@ -228,7 +228,7 @@ function renderHop(series, originalVersion) {
   let body;
   if (hop.actions.length === 0) {
     body = `<div class="hop dropin">
-      Drop-in hop: no code changes needed for ${series}.x &rarr; ${targetRelease}. Just continue.
+      You can upgrade directly to version ${targetRelease}. No changes needed.
     </div>`;
   } else {
     body = `<div class="hop">
@@ -246,7 +246,7 @@ function renderHop(series, originalVersion) {
   }
 
   const nextLabel = hop.to === LATEST ? `Finish at ${targetRelease} &rarr;` : `Continue to ${targetRelease} &rarr;`;
-  const reminder = hop.actions.length === 0 ?  `<p class="reminder">You can upgrade directly to the version showing on the "Continue" button. No changes needed.</p>` :
+  const reminder = hop.actions.length === 0 ?  `<p class="reminder">You can upgrade directly to version ${targetRelease}. No changes needed.</p>` :
     `<p class="reminder">If there are actions listed above, please ensure they are fixed before clicking "Continue".</p>`;
   app.innerHTML = `
     <p class="progress">Upgrading from ${series}.x &rarr; ${targetRelease}</p>
@@ -288,7 +288,7 @@ function renderHop(series, originalVersion) {
       } else if (yesAnswers.length === 0) {
         // All answered, all "No" — nothing applies, upgrade is direct.
         reminderEl.style.display = "";
-        reminderEl.textContent = 'You can upgrade directly to the version showing on the "Continue" button. No changes needed.';
+        reminderEl.textContent = `You can upgrade directly to version ${targetRelease}. No changes needed.`;
       } else {
         // All answered, at least one "Yes" — actions must be applied first.
         reminderEl.style.display = "";
@@ -342,17 +342,21 @@ function renderDone(originalVersion) {
   const fullVersion  = LATEST_RELEASE || LATEST;
   const seriesOnly   = LATEST;
   const summaryHTML  = renderSummary();
+  const introHTML    = decisions.length === 0
+    ? `<p><strong>You can upgrade directly to version ${fullVersion}. No changes needed.</strong></p>
+       <p>You're already in the ${seriesOnly}.x series, so the upgrade is just a Gemfile bump.</p>`
+    : `<p>You've completed all the upgrade steps from ${escapeHTML(originalVersion || "your current version")} to ${fullVersion} (the latest version in the ${seriesOnly}.x series).</p>`;
 
   app.innerHTML = `
     <div class="done">
       <h2>You're done</h2>
-      <p>You've walked all hops from ${escapeHTML(originalVersion || "your current version")} to ${fullVersion} (the latest patch in the ${seriesOnly}.x series).</p>
+      ${introHTML}
       ${summaryHTML}
       <p>Update your <code>Gemfile</code> to:</p>
       <pre><code>gem 'smarter_csv', '~&gt; ${seriesOnly}.0'</code></pre>
       <p>Then run:</p>
       <pre><code>bundle update smarter_csv</code></pre>
-      <p>After that, run your test suite. If anything behaves unexpectedly, click "Start over" and walk back through the hops to find the migration step you might have missed.</p>
+      <p>After that, run your test suite. If anything behaves unexpectedly, click "Start over" and walk back through the steps to find the migration item you might have missed.</p>
       <p class="muted">Questions? Open an issue at <a href="https://github.com/tilo/smarter_csv/issues">github.com/tilo/smarter_csv/issues</a>.</p>
       <p><button id="restart">Start over</button></p>
     </div>
@@ -366,17 +370,17 @@ function renderSummary() {
   const matchedCount = decisions.reduce((n, d) => n + d.matched.length, 0);
   const dropInCount  = decisions.filter(d => d.dropIn).length;
   const intro = matchedCount === 0
-    ? `<p>Good news: <strong>none of the per-hop conditions applied to your code</strong> across the ${decisions.length} hops you walked${dropInCount ? ` (${dropInCount} were drop-in hops)` : ""}. The upgrade should be a clean Gemfile bump.</p>`
-    : `<p>Across ${decisions.length} hops you walked, <strong>${matchedCount} migration step${matchedCount === 1 ? "" : "s"} apply</strong> to your code. Review and apply them before running <code>bundle update</code>:</p>`;
+    ? `<p>Good news &mdash; <strong>none of the conditions applied to your code</strong>. The upgrade is just a Gemfile bump.</p>`
+    : `<p>Apply the following <strong>${matchedCount} change${matchedCount === 1 ? "" : "s"}</strong> to your code, then run <code>bundle update</code>:</p>`;
 
   const list = decisions.map(d => {
     const targetRelease = latestReleaseFor(d.to);
     const heading = `<p class="summary-hop-heading"><strong>${d.from}.x &rarr; ${targetRelease}</strong></p>`;
     if (d.dropIn) {
-      return `<div class="summary-hop">${heading}<p class="muted">Drop-in hop &mdash; no code changes needed.</p></div>`;
+      return `<div class="summary-hop">${heading}<p class="muted">No code changes needed for this step.</p></div>`;
     }
     if (d.matched.length === 0) {
-      return `<div class="summary-hop">${heading}<p class="muted">None of the conditions on this hop applied to your code.</p></div>`;
+      return `<div class="summary-hop">${heading}<p class="muted">None of the conditions in this step applied to your code.</p></div>`;
     }
     const items = d.matched.map(a => `<li><strong>If</strong> ${a["if"]}<br>&rarr; ${a.then}</li>`).join("");
     return `<div class="summary-hop">${heading}<ul>${items}</ul></div>`;

data/ext/smarter_csv/smarter_csv.c

@@ -7,6 +7,14 @@
 #include <stdlib.h>
 #include <errno.h>
 
+#ifdef __ARM_NEON
+  #include <arm_neon.h>
+#elif defined(__SSE2__)
+  #include <immintrin.h>
+#endif
+
+#include "vendor/eisel_lemire.h" /* Eisel-Lemire decimal->double, correctly rounded (fast_float) */
+
 #ifndef bool
   #define bool int
   #define false ((bool)0)
@@ -41,6 +49,8 @@ static ID id_only, id_except, id_quote_boundary;
 static ID id_only_headers, id_except_headers, id_keep_cols, id_strict;
 static ID id_keep_bitmap, id_keep_extra_cols, id_early_exit_after_sym;
 static ID id_backslash, id_standard;
+static ID id_decimal_precision, id_float, id_bigdecimal;
+static ID id_BigDecimal; /* the Kernel#BigDecimal() method (require 'bigdecimal' done in Ruby) */
 
 /* ================================================================================
  * ParseContext — wraps all per-file parse options as a GC-managed TypedData object.
@@ -70,6 +80,9 @@ typedef struct {
   /* Numeric conversion: 0=off, 1=all, 2=only listed keys, 3=except listed keys */
   int  numeric_mode;
 
+  /* Decimal handling: 0=float, 1=auto (BigDecimal above 16 sig digits), 2=bigdecimal */
+  int  decimal_precision;
+
   /* Column filter bitmap (xmalloc'd; NULL when no filtering active) */
   bool *keep_bitmap;
   long  keep_bitmap_len;
@@ -133,6 +146,51 @@ static const rb_data_type_t parse_context_type = {
   RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_WB_PROTECTED
 };
 
+/* Scan [p, end) for the first `quote` char or backslash; returns a pointer to it,
+ * or `end` if neither occurs. NEON (arm64) or SSE2 (x86-64) processes 16 bytes per
+ * iteration; scalar fallback elsewhere. Ported from smarter_json's fj_scan_str.
+ *
+ * Used by the quoted-field slow path in :backslash escaping mode, where the only bytes
+ * that can change parser state inside a quoted field are the quote char (closing /
+ * doubled) and the backslash (escape). Bulk-skipping the plain content between them
+ * keeps the byte-by-byte state machine's behavior but avoids stepping every byte.
+ * In RFC mode the slow path uses a plain memchr-to-quote instead (only one byte class
+ * matters there), so this two-class scan is reserved for backslash mode. */
+static inline const char *scan_quote_or_backslash(const char *p, const char *end, char quote) {
+#ifdef __ARM_NEON
+  const uint8x16_t vq  = vdupq_n_u8((uint8_t)quote);
+  const uint8x16_t vbs = vdupq_n_u8((uint8_t)'\\');
+  while (p + 16 <= end) {
+    uint8x16_t chunk = vld1q_u8((const uint8_t *)p);
+    uint8x16_t m     = vorrq_u8(vceqq_u8(chunk, vq), vceqq_u8(chunk, vbs));
+    /* movemask emulation (Oj's technique): pack to 4 bits/byte, then ctz/4. */
+    uint8x8_t  res   = vshrn_n_u16(vreinterpretq_u16_u8(m), 4);
+    uint64_t   mask  = vget_lane_u64(vreinterpret_u64_u8(res), 0);
+    if (__builtin_expect(mask != 0, 0)) {  /* most 16-byte chunks contain neither */
+      mask &= 0x8888888888888888ull;
+      return p + (__builtin_ctzll(mask) >> 2);
+    }
+    p += 16;
+  }
+#elif defined(__SSE2__)
+  const __m128i vq  = _mm_set1_epi8(quote);
+  const __m128i vbs = _mm_set1_epi8('\\');
+  while (p + 16 <= end) {
+    __m128i chunk = _mm_loadu_si128((const __m128i *)p);
+    __m128i m     = _mm_or_si128(_mm_cmpeq_epi8(chunk, vq), _mm_cmpeq_epi8(chunk, vbs));
+    int     mask  = _mm_movemask_epi8(m);  /* one bit per byte that matched */
+    if (__builtin_expect(mask != 0, 0)) {  /* most 16-byte chunks contain neither */
+      return p + __builtin_ctz(mask);
+    }
+    p += 16;
+  }
+#endif
+  for (; p < end; p++) {
+    if (*p == quote || *p == '\\') return p;
+  }
+  return end;
+}
+
 static VALUE unescape_quotes(char *str, long len, char quote_char, rb_encoding *encoding) {
   // Fast path: scan for any doubled quote pair. If none present, the field has
   // nothing to unescape — emit it directly via rb_enc_str_new and skip the
@@ -386,6 +444,20 @@ static VALUE rb_parse_csv_line(VALUE self, VALUE line, VALUE col_sep, VALUE quot
       backslash_count = 0;
       field_started = false;  // reset for next field
     } else {
+      /* Backslash mode: NEON scan-ahead to the next quote OR backslash (Opt #7).
+       * Inside a quoted field the only state-changing bytes are the quote char and the
+       * backslash; bulk-skip the plain content between them. Skipped bytes are plain
+       * content, which the byte-by-byte loop resets backslash_count to 0 on, so reset
+       * it here whenever we actually move p. */
+      if (allow_escaped_quotes && in_quotes) {
+        const char *hit = scan_quote_or_backslash(p, endP, quote_char_val);
+        if (hit != p) {
+          backslash_count = 0;
+          p = (char *)hit;
+          if (p == endP) continue;  /* no quote/backslash before end → unclosed */
+        }
+      }
+
       if (allow_escaped_quotes && *p == '\\') {
         backslash_count++;
         if (__builtin_expect(quote_boundary_standard, 1) && !in_quotes) field_started = true;
@@ -525,47 +597,101 @@ static inline VALUE get_key_for_index(long index, VALUE headers, long headers_le
  * Handles overflow: if strtol overflows (ERANGE), falls back to rb_cstr_to_inum
  * which produces a Ruby Bignum.
  */
-static inline VALUE try_numeric_conversion(char *trim_start, long trimmed_len) {
-  // Quick pre-check: first char must be digit, +, -, or .
-  char first = trim_start[0];
-  if (!((first >= '0' && first <= '9') || first == '+' || first == '-' || first == '.')) {
+static inline VALUE try_numeric_conversion(char *s, long n, int decimal_precision) {
+  // Quick pre-check: first char must be a digit or a sign.
+  char first = s[0];
+  if (!((first >= '0' && first <= '9') || first == '+' || first == '-')) {
     return Qundef;
   }
 
-  // Need null-terminated string for strtol/strtod; use stack buffer for typical fields
-  if (trimmed_len >= 63) return Qundef;  // very long fields are unlikely to be simple numbers
-
-  char num_buf[64];
-  memcpy(num_buf, trim_start, trimmed_len);
-  num_buf[trimmed_len] = '\0';
-
-  char *endptr;
-
-  // Try integer first (most common numeric type in CSV)
-  // Don't try integer if field starts with '.' (e.g., ".5")
-  if (first != '.') {
-    errno = 0;
-    long int_val = strtol(num_buf, &endptr, 10);
-    if (endptr == num_buf + trimmed_len) {
-      // Entire string was consumed → valid integer
-      if (errno == ERANGE) {
-        // Overflow: fall back to Ruby Bignum
-        return rb_cstr_to_inum(num_buf, 10, false);
+  /* Single pass: validate the token against the same grammar as the Ruby path's
+   * NUMERIC_REGEX = /\A[+-]?\d+(?:\.\d+)?(?:[eE][+-]?\d+)?\z/ and, in the same pass,
+   * extract everything the fast paths need:
+   *   - mantissa value m10 (exact for <= 18 digits; `overflow` flags beyond)
+   *   - significant-digit count `sig` (leading zeros excluded; matches the Ruby
+   *     significant_digits helper / Oj dec_cnt) — drives the :auto Float/BigDecimal split
+   *   - base-10 exponent e10 (from the fraction length and any explicit exponent)
+   * Anything the grammar rejects returns Qundef (stays a String), keeping the C and
+   * Ruby paths byte-identical on what does and does not convert. */
+  long i = 0;
+  int neg = 0;
+  if (s[i] == '+' || s[i] == '-') { neg = (s[i] == '-'); i++; }
+
+  uint64_t m10 = 0;
+  int  m10digits = 0;       /* mantissa digits accumulated into m10 (capped at 19) */
+  long sig = 0;             /* significant digits (leading zeros excluded) */
+  int  sig_started = 0;
+  bool overflow = false;
+  long int_digits = 0, frac_digits = 0;
+  bool seen_dot = false, seen_exp = false, any_digit = false, exp_any = false;
+  int64_t exp_val = 0; int exp_neg = 0;
+
+  for (; i < n; i++) {
+    char c = s[i];
+    if (c >= '0' && c <= '9') {
+      any_digit = true;
+      if (!seen_exp) {
+        if (seen_dot) frac_digits++; else int_digits++;
+        if (sig_started) sig++;
+        else if (c != '0') { sig_started = 1; sig = 1; }
+        if (m10digits < 19) { m10 = m10 * 10 + (uint64_t)(c - '0'); m10digits++; }
+        else overflow = true;
+      } else {
+        exp_any = true;
+        exp_val = exp_val * 10 + (c - '0');
+        if (exp_val > 1000000) overflow = true; /* extreme exponent → strtod fallback */
       }
-      return LONG2NUM(int_val);
+    } else if (c == '.' && !seen_dot && !seen_exp) {
+      seen_dot = true;
+    } else if ((c == 'e' || c == 'E') && !seen_exp && any_digit) {
+      seen_exp = true;
+      if (i + 1 < n && (s[i + 1] == '+' || s[i + 1] == '-')) { exp_neg = (s[i + 1] == '-'); i++; }
+    } else {
+      return Qundef; /* invalid char for a number → not numeric */
     }
   }
 
-  // Try float (only if contains '.')
-  if (memchr(num_buf, '.', trimmed_len)) {
-    errno = 0;
-    double float_val = strtod(num_buf, &endptr);
-    if (endptr == num_buf + trimmed_len && errno != ERANGE) {
-      return DBL2NUM(float_val);
+  /* Enforce NUMERIC_REGEX exactly: an integer part is required; a dot requires a
+   * fraction digit; an exponent requires an exponent digit. */
+  if (int_digits == 0) return Qundef;
+  if (seen_dot && frac_digits == 0) return Qundef;
+  if (seen_exp && !exp_any) return Qundef;
+
+  bool is_decimal = seen_dot || seen_exp;
+
+  if (!is_decimal) {
+    /* Integer. Fast path when it fits in a long; otherwise a Ruby Integer/Bignum. */
+    if (!overflow && m10digits <= 18) {
+      long v = (long)m10;
+      return LONG2NUM(neg ? -v : v);
     }
+    VALUE str = rb_str_new(s, n);
+    return rb_cstr_to_inum(RSTRING_PTR(str), 10, false);
   }
 
-  return Qundef;  // not numeric
+  /* Decimal (has a '.' or an exponent) — honor decimal_precision. 0=float, 1=auto, 2=bigdecimal */
+  if (decimal_precision == 2 || (decimal_precision == 1 && sig > 16)) {
+    VALUE str = rb_str_new(s, n);
+    return rb_funcall(rb_cObject, id_BigDecimal, 1, str);
+  }
+
+  /* Float. base-10 exponent = explicit exponent minus the fraction length. */
+  int64_t e10 = (exp_neg ? -exp_val : exp_val) - (int64_t)frac_digits;
+  double d;
+  if (!overflow && m10digits >= 1 && m10digits <= 19 && ((long)m10digits + e10) >= -307) {
+    /* Eisel-Lemire is correctly-rounded for any nonzero mantissa that fits exactly in a
+     * uint64 — i.e. up to 19 significant digits (the max 19-digit value ~1.0e19 is below
+     * UINT64_MAX ~1.8e19). Verified bit-for-bit vs the stdlib over 1..19-digit ties. */
+    d = (m10 == 0) ? (neg ? -0.0 : 0.0) : fj_eisel_lemire_s2d(e10, m10, neg);
+  } else {
+    /* >19 digits / extreme or subnormal exponent: fall back to Ruby's own correctly-rounded
+     * strtod (rb_cstr_to_dbl) — the exact conversion String#to_f uses — so the C path and the
+     * Ruby path produce the identical double on every platform, not just where the system
+     * strtod happens to be correctly rounded. The token is pre-validated, so badcheck=0. */
+    VALUE str = rb_str_new(s, n);
+    d = rb_cstr_to_dbl(RSTRING_PTR(str), 0);
+  }
+  return DBL2NUM(d);
 }
 
 /*
@@ -614,6 +740,7 @@ typedef struct {
   long headers_len;
   long hash_capa;           // Pre-computed capacity for lazy hash allocation
   int numeric_mode;         // 0=off, 1=all, 2=only, 3=except
+  int decimal_precision;    // 0=float, 1=auto (BigDecimal above 16 sig digits), 2=bigdecimal
   bool remove_empty_values;
   bool remove_zero_values;
 } field_transform_opts;
@@ -705,7 +832,7 @@ static inline __attribute__((always_inline)) bool insert_field_into_hash(
                       (opts->numeric_mode == 2 && rb_ary_includes(opts->numeric_keys, key) == Qtrue) ||
                       (opts->numeric_mode == 3 && rb_ary_includes(opts->numeric_keys, key) != Qtrue);
     if (do_convert) {
-      VALUE numeric = try_numeric_conversion(trim_start, trimmed_len);
+      VALUE numeric = try_numeric_conversion(trim_start, trimmed_len, opts->decimal_precision);
       if (numeric != Qundef) {
         ensure_hash_allocated(opts);
         rb_hash_aset(opts->hash, key, numeric);
@@ -752,6 +879,18 @@ void parse_numeric_option(VALUE options_hash, int *out_mode, VALUE *out_keys) {
   }
 }
 
+/* Read decimal_precision into 0=float, 1=auto, 2=bigdecimal. Default :auto (1).
+ * The option is validated and coerced to a symbol on the Ruby side before we get here. */
+static inline int parse_decimal_precision(VALUE options_hash) {
+  VALUE v = rb_hash_aref(options_hash, ID2SYM(id_decimal_precision));
+  if (RB_TYPE_P(v, T_SYMBOL)) {
+    ID s = SYM2ID(v);
+    if (s == id_float) return 0;
+    if (s == id_bigdecimal) return 2;
+  }
+  return 1; // :auto (also the default when unset)
+}
+
 /*
  * ================================================================================
  * rb_parse_line_to_hash - Parse CSV line directly into a Ruby Hash
@@ -826,6 +965,7 @@ __attribute__((hot)) static VALUE rb_parse_line_to_hash(VALUE self, VALUE line,
   int numeric_mode = 0;
   VALUE numeric_keys = Qnil;
   parse_numeric_option(options_hash, &numeric_mode, &numeric_keys);
+  int decimal_precision = parse_decimal_precision(options_hash);
 
   // quote_escaping and quote_boundary are only needed in Section 5 (quoted/slow path).
   // They are declared here as forward declarations so Section 5 can set them lazily.
@@ -990,6 +1130,7 @@ __attribute__((hot)) static VALUE rb_parse_line_to_hash(VALUE self, VALUE line,
     .headers_len = headers_len,
     .hash_capa = hash_size,
     .numeric_mode = numeric_mode,
+    .decimal_precision = decimal_precision,
     .remove_empty_values = remove_empty_values,
     .remove_zero_values = remove_zero_values,
   };
@@ -1160,6 +1301,20 @@ __attribute__((hot)) static VALUE rb_parse_line_to_hash(VALUE self, VALUE line,
           p = next_quote;  /* jump to quote char; fall through to quote-handling code */
         }
 
+        /* Backslash mode: NEON scan-ahead to the next quote OR backslash (Opt #7).
+         * The RFC memchr skip above only matters for one byte class; with escaping on
+         * a backslash also changes state, so scan for both. Skipped bytes are plain
+         * content (the byte-by-byte loop resets backslash_count to 0 on them), so reset
+         * it here whenever we actually move p. */
+        if (allow_escaped_quotes && in_quotes) {
+          const char *hit = scan_quote_or_backslash(p, endP, quote_char_val);
+          if (hit != p) {
+            backslash_count = 0;
+            p = (char *)hit;
+            if (p == endP) continue;  /* no quote/backslash before end → unclosed */
+          }
+        }
+
         if (allow_escaped_quotes && *p == '\\') {
           // Count consecutive backslashes for escape sequence detection
           backslash_count++;
@@ -1354,6 +1509,7 @@ __attribute__((cold)) static VALUE rb_new_parse_context(VALUE self, VALUE header
 
   /* Numeric conversion */
   parse_numeric_option(options_hash, &ctx->numeric_mode, &ctx->numeric_keys);
+  ctx->decimal_precision = parse_decimal_precision(options_hash);
 
   /* quote_escaping → allow_escaped_quotes */
   VALUE quote_escaping_val = rb_hash_aref(options_hash, ID2SYM(id_quote_escaping));
@@ -1474,6 +1630,7 @@ __attribute__((hot)) static VALUE rb_parse_line_to_hash_ctx(VALUE self, VALUE li
   bool remove_empty_values = ctx->remove_empty_values;
   bool remove_zero_values  = ctx->remove_zero_values;
   int  numeric_mode        = ctx->numeric_mode;
+  int  decimal_precision   = ctx->decimal_precision;
   VALUE numeric_keys       = ctx->numeric_keys;
   bool *keep_bitmap         = ctx->keep_bitmap;
   /* keep_bitmap is cached in the context (xmalloc'd once at construction, sized to the header count
@@ -1525,6 +1682,7 @@ __attribute__((hot)) static VALUE rb_parse_line_to_hash_ctx(VALUE self, VALUE li
     .headers_len       = headers_len,
     .hash_capa         = hash_size,
     .numeric_mode      = numeric_mode,
+    .decimal_precision = decimal_precision,
     .remove_empty_values = remove_empty_values,
     .remove_zero_values  = remove_zero_values,
   };
@@ -1654,6 +1812,16 @@ __attribute__((hot)) static VALUE rb_parse_line_to_hash_ctx(VALUE self, VALUE li
           p = next_quote;  /* fall through to quote-handling code */
         }
 
+        /* Backslash mode: NEON scan-ahead to the next quote OR backslash (Opt #7). */
+        if (allow_escaped_quotes && in_quotes) {
+          const char *hit = scan_quote_or_backslash(p, endP, quote_char_val);
+          if (hit != p) {
+            backslash_count = 0;
+            p = (char *)hit;
+            if (p == endP) continue;  /* no quote/backslash before end → unclosed */
+          }
+        }
+
         if (allow_escaped_quotes && *p == '\\') {
           backslash_count++;
           if (__builtin_expect(quote_boundary_standard, 1) && !in_quotes) field_started = true;
@@ -1878,6 +2046,10 @@ void Init_smarter_csv(void) {
   id_strict             = rb_intern("strict");
   id_backslash      = rb_intern("backslash");
   id_standard       = rb_intern("standard");
+  id_decimal_precision = rb_intern("decimal_precision");
+  id_float          = rb_intern("float");
+  id_bigdecimal     = rb_intern("bigdecimal");
+  id_BigDecimal     = rb_intern("BigDecimal"); /* Kernel#BigDecimal(); 'bigdecimal' is required in lib/smarter_csv.rb */
 
   rb_define_module_function(Parser, "parse_csv_line_c", rb_parse_csv_line, 9);
   rb_define_module_function(Parser, "count_quote_chars_c", rb_count_quote_chars, 4);