PyPI - pymodab - Versions diffs - 1.1.2__tar.gz → 1.1.3__tar.gz - Mend

pymodab 1.1.2tar.gz → 1.1.3tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (20) hide show

{pymodab-1.1.2/src/pymodab.egg-info → pymodab-1.1.3}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pymodab
-Version: 1.1.2
+Version: 1.1.3
 Summary: Fast and robust root-finding library using the Modified Anderson-Bjork method
 Author: Ganchovski, N.; Smith, O.; Rackauckas, C.; Tomov, L.; Traykov, A.
 License-Expression: MIT
@@ -90,6 +90,8 @@ Version 1.1.1 separates the bisection and Anderson-Björck steps in the solver l
 Version 1.1.2 streamlines the solver loop. The bracket width and the difference of the true residuals are computed once per step and reused. The exits on a zero and on a `NaN` residual share a single test on the hot path. The best true residual of the bracket, which the fallback test needs, is now computed only when a secant step fails to shrink the bracket fast enough, instead of on every secant step, and the test is written as `2·|y3| >= ymin` instead of `|y3| >= 0.5·ymin`. The 101 benchmark problems give the same roots and evaluation counts as with 1.1.1.
+Version 1.1.3 replaces the capped residual exception of 1.1.1 with a stricter test and removes the `MAX_RESIDUAL_STEPS` counter. A secant step that does not shrink the bracket fast enough is now kept only if it reduces the residual more than 42 times, that is `42·|y3| < ymin`, where `ymin` is the smaller true residual of the bracket. The factor comes from 2^(1074/200) ≈ 41.4: at this rate a residual that starts near 1 underflows to zero within the default limit of 200 iterations, because the smallest positive double is 2^-1074. A search that stays on secant steps this way therefore ends by itself, without a counter. The factor is a fixed constant and does not follow `max_iter`, so raising the limit does not weaken the test. On the benchmark, problem f101 now needs 65 evaluations instead of 76, and the other 100 problems give the same roots and evaluation counts as with 1.1.2; all 101 need 1960 evaluations in total instead of 1971.
 For maximum speed, pass a compiled function, e.g. from [numba](https://numba.pydata.org/). It is then called from C with no Python overhead:
 ```python
@@ -142,27 +144,27 @@ The modAB algorithm is benchmarked against the available algorithms in Python/Sc
   Func|   bisect|   brentq|   brenth|   ridder|   chandr| cybrentq| modAB_ct| <mark>**modAB_SG**</mark>|
 ----- | ------: | ------: | ------: | ------: | ------: | ------: | ------: | ------: |
-   SUM|     4883|     2810|     2769|     3425|     2112|     2810|     1993| <mark>**1971**</mark>|
-   AVG|       48|       28|       27|       34|       21|       28|       20| <mark>**20**</mark>|
+   SUM|     4883|     2810|     2769|     3425|     2112|     2810|     1993| <mark>**1960**</mark>|
+   AVG|       48|       28|       27|       34|       21|       28|       20| <mark>**19**</mark>|
 MEDIAN|       49|       13|       12|       16|       12|       13|       13| <mark>**13**</mark>|
    MIN|        3|        4|        4|        3|        3|        4|        3| <mark>**3**</mark>|
-   MAX|       53|      102|      102|      202|       70|      102|       78| <mark>**76**</mark>|
-FACTOR|   2.477x|   1.426x|   1.405x|   1.738x|   1.072x|   1.426x|   1.011x| <mark>**1.000x**</mark>|
+   MAX|       53|      102|      102|      202|       70|      102|       78| <mark>**65**</mark>|
+FACTOR|   2.491x|   1.434x|   1.413x|   1.747x|   1.078x|   1.434x|   1.017x| <mark>**1.000x**</mark>|
 #### Execution times  (ms per problem, 200 iterations)
   Func|   bisect|   brentq|   brenth|   ridder|   chandr| cybrentq| modAB_ct| <mark>**modAB_SG**</mark>|
 ----- | ------: | ------: | ------: | ------: | ------: | ------: | ------: | ------: |
-   SUM|  1497.81|   888.81|   878.13|  1103.39| 62435.64|   118.77|   176.71|<mark>**90.70**</mark>|
-   AVG|  14.8298|   8.8001|   8.6944|  10.9246| 618.1747|   1.1759|   1.7496|<mark>**0.8981**</mark>|
-MEDIAN|  14.1046|   4.5237|   4.4239|   5.5108| 366.5141|   0.6629|   1.3385|<mark>**0.6881**</mark>|
-   MIN|   1.6523|   1.6333|   1.8243|   1.4128|  68.0820|   0.1926|   0.5485|<mark>**0.1625**</mark>|
-   MAX|  39.1626|  36.7848|  31.6445| 121.3956| 2159.4610|   5.8503|   5.9590|<mark>**3.2718**</mark>|
-FACTOR|  16.513x|   9.799x|   9.681x|  12.165x| 688.344x|   1.309x|   1.948x|<mark>**1.000x**</mark>|
+   SUM|  1535.42|   931.31|   878.87|  1047.95| 58524.74|   115.55|   175.60| <mark>**91.23**</mark>|
+   AVG|  15.2021|   9.2209|   8.7017|  10.3758| 579.4528|   1.1440|   1.7386| <mark>**0.9033**</mark>|
+MEDIAN|  14.7569|   5.1910|   4.9383|   6.4764| 345.7968|   0.5879|   1.3410| <mark>**0.6511**</mark>|
+   MIN|   1.7242|   1.8918|   1.8852|   2.0780|  73.8264|   0.2071|   0.5655| <mark>**0.1677**</mark>|
+   MAX|  70.1038|  32.1594|  41.2759|  53.5572|1979.9811|   6.4304|   5.1045| <mark>**3.8393**</mark>|
+FACTOR|  16.829x|  10.208x|   9.633x|  11.486x| 641.480x|   1.266x|   1.925x| <mark>**1.000x**</mark>|
 #### Notes:
-Last Run on: 06.10.2026
+Last Run on: 01.11.2026
 Intel(R) Core(TM) i7-1065G7 CPU @ 1.30GHz (1.50 GHz) with 16.0 GB RAM
 Windows 11 Home
 Python Version: 3.14.7
@@ -170,4 +172,4 @@ numpy Version: 2.4.6
 scipy Version: 1.18.0
 cybrentq Version: 0.1.5 - by Gledis Caushaj (https://github.com/gledi-ai/cybrentq)
 modAB_ct: pymodab version 1.0.5 from PyPI - the previous implementation with ctypes
-modAB_SG: pymodab version 1.1.2 - the latest implementation of the safeguarded modab as native C extension
+modAB_SG: pymodab version 1.1.3 - the latest implementation of the safeguarded modab as native C extension

{pymodab-1.1.2 → pymodab-1.1.3}/README.md RENAMED Viewed

@@ -60,6 +60,8 @@ Version 1.1.1 separates the bisection and Anderson-Björck steps in the solver l
 Version 1.1.2 streamlines the solver loop. The bracket width and the difference of the true residuals are computed once per step and reused. The exits on a zero and on a `NaN` residual share a single test on the hot path. The best true residual of the bracket, which the fallback test needs, is now computed only when a secant step fails to shrink the bracket fast enough, instead of on every secant step, and the test is written as `2·|y3| >= ymin` instead of `|y3| >= 0.5·ymin`. The 101 benchmark problems give the same roots and evaluation counts as with 1.1.1.
+Version 1.1.3 replaces the capped residual exception of 1.1.1 with a stricter test and removes the `MAX_RESIDUAL_STEPS` counter. A secant step that does not shrink the bracket fast enough is now kept only if it reduces the residual more than 42 times, that is `42·|y3| < ymin`, where `ymin` is the smaller true residual of the bracket. The factor comes from 2^(1074/200) ≈ 41.4: at this rate a residual that starts near 1 underflows to zero within the default limit of 200 iterations, because the smallest positive double is 2^-1074. A search that stays on secant steps this way therefore ends by itself, without a counter. The factor is a fixed constant and does not follow `max_iter`, so raising the limit does not weaken the test. On the benchmark, problem f101 now needs 65 evaluations instead of 76, and the other 100 problems give the same roots and evaluation counts as with 1.1.2; all 101 need 1960 evaluations in total instead of 1971.
 For maximum speed, pass a compiled function, e.g. from [numba](https://numba.pydata.org/). It is then called from C with no Python overhead:
 ```python
@@ -112,27 +114,27 @@ The modAB algorithm is benchmarked against the available algorithms in Python/Sc
   Func|   bisect|   brentq|   brenth|   ridder|   chandr| cybrentq| modAB_ct| <mark>**modAB_SG**</mark>|
 ----- | ------: | ------: | ------: | ------: | ------: | ------: | ------: | ------: |
-   SUM|     4883|     2810|     2769|     3425|     2112|     2810|     1993| <mark>**1971**</mark>|
-   AVG|       48|       28|       27|       34|       21|       28|       20| <mark>**20**</mark>|
+   SUM|     4883|     2810|     2769|     3425|     2112|     2810|     1993| <mark>**1960**</mark>|
+   AVG|       48|       28|       27|       34|       21|       28|       20| <mark>**19**</mark>|
 MEDIAN|       49|       13|       12|       16|       12|       13|       13| <mark>**13**</mark>|
    MIN|        3|        4|        4|        3|        3|        4|        3| <mark>**3**</mark>|
-   MAX|       53|      102|      102|      202|       70|      102|       78| <mark>**76**</mark>|
-FACTOR|   2.477x|   1.426x|   1.405x|   1.738x|   1.072x|   1.426x|   1.011x| <mark>**1.000x**</mark>|
+   MAX|       53|      102|      102|      202|       70|      102|       78| <mark>**65**</mark>|
+FACTOR|   2.491x|   1.434x|   1.413x|   1.747x|   1.078x|   1.434x|   1.017x| <mark>**1.000x**</mark>|
 #### Execution times  (ms per problem, 200 iterations)
   Func|   bisect|   brentq|   brenth|   ridder|   chandr| cybrentq| modAB_ct| <mark>**modAB_SG**</mark>|
 ----- | ------: | ------: | ------: | ------: | ------: | ------: | ------: | ------: |
-   SUM|  1497.81|   888.81|   878.13|  1103.39| 62435.64|   118.77|   176.71|<mark>**90.70**</mark>|
-   AVG|  14.8298|   8.8001|   8.6944|  10.9246| 618.1747|   1.1759|   1.7496|<mark>**0.8981**</mark>|
-MEDIAN|  14.1046|   4.5237|   4.4239|   5.5108| 366.5141|   0.6629|   1.3385|<mark>**0.6881**</mark>|
-   MIN|   1.6523|   1.6333|   1.8243|   1.4128|  68.0820|   0.1926|   0.5485|<mark>**0.1625**</mark>|
-   MAX|  39.1626|  36.7848|  31.6445| 121.3956| 2159.4610|   5.8503|   5.9590|<mark>**3.2718**</mark>|
-FACTOR|  16.513x|   9.799x|   9.681x|  12.165x| 688.344x|   1.309x|   1.948x|<mark>**1.000x**</mark>|
+   SUM|  1535.42|   931.31|   878.87|  1047.95| 58524.74|   115.55|   175.60| <mark>**91.23**</mark>|
+   AVG|  15.2021|   9.2209|   8.7017|  10.3758| 579.4528|   1.1440|   1.7386| <mark>**0.9033**</mark>|
+MEDIAN|  14.7569|   5.1910|   4.9383|   6.4764| 345.7968|   0.5879|   1.3410| <mark>**0.6511**</mark>|
+   MIN|   1.7242|   1.8918|   1.8852|   2.0780|  73.8264|   0.2071|   0.5655| <mark>**0.1677**</mark>|
+   MAX|  70.1038|  32.1594|  41.2759|  53.5572|1979.9811|   6.4304|   5.1045| <mark>**3.8393**</mark>|
+FACTOR|  16.829x|  10.208x|   9.633x|  11.486x| 641.480x|   1.266x|   1.925x| <mark>**1.000x**</mark>|
 #### Notes:
-Last Run on: 06.10.2026
+Last Run on: 01.11.2026
 Intel(R) Core(TM) i7-1065G7 CPU @ 1.30GHz (1.50 GHz) with 16.0 GB RAM
 Windows 11 Home
 Python Version: 3.14.7
@@ -140,4 +142,4 @@ numpy Version: 2.4.6
 scipy Version: 1.18.0
 cybrentq Version: 0.1.5 - by Gledis Caushaj (https://github.com/gledi-ai/cybrentq)
 modAB_ct: pymodab version 1.0.5 from PyPI - the previous implementation with ctypes
-modAB_SG: pymodab version 1.1.2 - the latest implementation of the safeguarded modab as native C extension
+modAB_SG: pymodab version 1.1.3 - the latest implementation of the safeguarded modab as native C extension

{pymodab-1.1.2 → pymodab-1.1.3}/pyproject.toml RENAMED Viewed

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "pymodab"
-version = "1.1.2"
+version = "1.1.3"
 description = "Fast and robust root-finding library using the Modified Anderson-Bjork method"
 readme = "README.md"
 license = "MIT"

{pymodab-1.1.2 → pymodab-1.1.3}/src/pymodab/ModAB.dll RENAMED Viewed

Binary file

{pymodab-1.1.2 → pymodab-1.1.3}/src/pymodab/__init__.py RENAMED Viewed

@@ -11,5 +11,5 @@ Preprints 2026, 2026032190. https://doi.org/10.20944/preprints202603.2190.v1
 from .modab import find_root, get_evaluation_count, NATIVE
-__version__ = "1.1.2"
+__version__ = "1.1.3"
 __all__ = ["find_root", "get_evaluation_count"]

{pymodab-1.1.2 → pymodab-1.1.3}/src/pymodab/_modab.c RENAMED Viewed

@@ -148,8 +148,6 @@ static double modab_core(eval_fn f, void *ctx, double x1, double x2,
     double threshold = x2 - x1;
     double f1 = y1, f2 = y2; /* True residuals, kept unmodified by A&B corrections */
     const double C = 2.0;
-    const int MAX_RESIDUAL_STEPS = 3; /* Max consecutive A&B steps kept by the residual test alone */
-    int residual_steps = 0;
     for (int i = 1; i <= maxIter; ++i) {
         double x3 = bisection ? safe_midpoint(x1, x2) : safe_secant(x1, y1, x2, y2);
         double dx = x2 - x1; /* Bracket width */
@@ -169,7 +167,6 @@ static double modab_core(eval_fn f, void *ctx, double x1, double x2,
                 if (fabs(ym - y3) < k * fabs(ym) + k * fabs(y3)) {
                     bisection = 0;
                     threshold = C * dx;
-                    residual_steps = 0;
                     y1 = f1; y2 = f2; /* A&B starts from the true residuals */
                 }
             }
@@ -213,17 +210,12 @@ static double modab_core(eval_fn f, void *ctx, double x1, double x2,
                     side = -1;
                 x2 = x3; f2 = y2 = y3;
             }
-            /* Fallback if AB fails to reduce the bracket width, unless it still halves the residual,
-               but for no more than MAX_RESIDUAL_STEPS consecutive steps */
-            if (x2 - x1 > threshold) {
-                double ymin = fmin(fabs(yl), fabs(yr)); /* Best true residual of the bracket */
-                if (residual_steps >= MAX_RESIDUAL_STEPS || 2.0 * fabs(y3) >= ymin) {
-                    bisection = 1;
-                    side = 0;
-                } else
-                    ++residual_steps;
-            } else
-                residual_steps = 0;
+            /* Fallback if AB fails to reduce the bracket width, unless it still shrinks the residual more than 42 times
+               (2^(1074/200) = 41.4: at this rate the residual underflows within 200 steps from |y| = 1) */
+            if (x2 - x1 > threshold && 42.0 * fabs(y3) >= fmin(fabs(yl), fabs(yr))) {
+                bisection = 1;
+                side = 0;
+            }
         }
     }
     return NAN;

pymodab-1.1.3/src/pymodab/libModAB.so ADDED Viewed

Binary file

{pymodab-1.1.2 → pymodab-1.1.3}/src/pymodab/libModAB_arm64.dylib RENAMED Viewed

Binary file

{pymodab-1.1.2 → pymodab-1.1.3}/src/pymodab/libModAB_x64.dylib RENAMED Viewed

Binary file

{pymodab-1.1.2 → pymodab-1.1.3/src/pymodab.egg-info}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pymodab
-Version: 1.1.2
+Version: 1.1.3
 Summary: Fast and robust root-finding library using the Modified Anderson-Bjork method
 Author: Ganchovski, N.; Smith, O.; Rackauckas, C.; Tomov, L.; Traykov, A.
 License-Expression: MIT
@@ -90,6 +90,8 @@ Version 1.1.1 separates the bisection and Anderson-Björck steps in the solver l
 Version 1.1.2 streamlines the solver loop. The bracket width and the difference of the true residuals are computed once per step and reused. The exits on a zero and on a `NaN` residual share a single test on the hot path. The best true residual of the bracket, which the fallback test needs, is now computed only when a secant step fails to shrink the bracket fast enough, instead of on every secant step, and the test is written as `2·|y3| >= ymin` instead of `|y3| >= 0.5·ymin`. The 101 benchmark problems give the same roots and evaluation counts as with 1.1.1.
+Version 1.1.3 replaces the capped residual exception of 1.1.1 with a stricter test and removes the `MAX_RESIDUAL_STEPS` counter. A secant step that does not shrink the bracket fast enough is now kept only if it reduces the residual more than 42 times, that is `42·|y3| < ymin`, where `ymin` is the smaller true residual of the bracket. The factor comes from 2^(1074/200) ≈ 41.4: at this rate a residual that starts near 1 underflows to zero within the default limit of 200 iterations, because the smallest positive double is 2^-1074. A search that stays on secant steps this way therefore ends by itself, without a counter. The factor is a fixed constant and does not follow `max_iter`, so raising the limit does not weaken the test. On the benchmark, problem f101 now needs 65 evaluations instead of 76, and the other 100 problems give the same roots and evaluation counts as with 1.1.2; all 101 need 1960 evaluations in total instead of 1971.
 For maximum speed, pass a compiled function, e.g. from [numba](https://numba.pydata.org/). It is then called from C with no Python overhead:
 ```python
@@ -142,27 +144,27 @@ The modAB algorithm is benchmarked against the available algorithms in Python/Sc
   Func|   bisect|   brentq|   brenth|   ridder|   chandr| cybrentq| modAB_ct| <mark>**modAB_SG**</mark>|
 ----- | ------: | ------: | ------: | ------: | ------: | ------: | ------: | ------: |
-   SUM|     4883|     2810|     2769|     3425|     2112|     2810|     1993| <mark>**1971**</mark>|
-   AVG|       48|       28|       27|       34|       21|       28|       20| <mark>**20**</mark>|
+   SUM|     4883|     2810|     2769|     3425|     2112|     2810|     1993| <mark>**1960**</mark>|
+   AVG|       48|       28|       27|       34|       21|       28|       20| <mark>**19**</mark>|
 MEDIAN|       49|       13|       12|       16|       12|       13|       13| <mark>**13**</mark>|
    MIN|        3|        4|        4|        3|        3|        4|        3| <mark>**3**</mark>|
-   MAX|       53|      102|      102|      202|       70|      102|       78| <mark>**76**</mark>|
-FACTOR|   2.477x|   1.426x|   1.405x|   1.738x|   1.072x|   1.426x|   1.011x| <mark>**1.000x**</mark>|
+   MAX|       53|      102|      102|      202|       70|      102|       78| <mark>**65**</mark>|
+FACTOR|   2.491x|   1.434x|   1.413x|   1.747x|   1.078x|   1.434x|   1.017x| <mark>**1.000x**</mark>|
 #### Execution times  (ms per problem, 200 iterations)
   Func|   bisect|   brentq|   brenth|   ridder|   chandr| cybrentq| modAB_ct| <mark>**modAB_SG**</mark>|
 ----- | ------: | ------: | ------: | ------: | ------: | ------: | ------: | ------: |
-   SUM|  1497.81|   888.81|   878.13|  1103.39| 62435.64|   118.77|   176.71|<mark>**90.70**</mark>|
-   AVG|  14.8298|   8.8001|   8.6944|  10.9246| 618.1747|   1.1759|   1.7496|<mark>**0.8981**</mark>|
-MEDIAN|  14.1046|   4.5237|   4.4239|   5.5108| 366.5141|   0.6629|   1.3385|<mark>**0.6881**</mark>|
-   MIN|   1.6523|   1.6333|   1.8243|   1.4128|  68.0820|   0.1926|   0.5485|<mark>**0.1625**</mark>|
-   MAX|  39.1626|  36.7848|  31.6445| 121.3956| 2159.4610|   5.8503|   5.9590|<mark>**3.2718**</mark>|
-FACTOR|  16.513x|   9.799x|   9.681x|  12.165x| 688.344x|   1.309x|   1.948x|<mark>**1.000x**</mark>|
+   SUM|  1535.42|   931.31|   878.87|  1047.95| 58524.74|   115.55|   175.60| <mark>**91.23**</mark>|
+   AVG|  15.2021|   9.2209|   8.7017|  10.3758| 579.4528|   1.1440|   1.7386| <mark>**0.9033**</mark>|
+MEDIAN|  14.7569|   5.1910|   4.9383|   6.4764| 345.7968|   0.5879|   1.3410| <mark>**0.6511**</mark>|
+   MIN|   1.7242|   1.8918|   1.8852|   2.0780|  73.8264|   0.2071|   0.5655| <mark>**0.1677**</mark>|
+   MAX|  70.1038|  32.1594|  41.2759|  53.5572|1979.9811|   6.4304|   5.1045| <mark>**3.8393**</mark>|
+FACTOR|  16.829x|  10.208x|   9.633x|  11.486x| 641.480x|   1.266x|   1.925x| <mark>**1.000x**</mark>|
 #### Notes:
-Last Run on: 06.10.2026
+Last Run on: 01.11.2026
 Intel(R) Core(TM) i7-1065G7 CPU @ 1.30GHz (1.50 GHz) with 16.0 GB RAM
 Windows 11 Home
 Python Version: 3.14.7
@@ -170,4 +172,4 @@ numpy Version: 2.4.6
 scipy Version: 1.18.0
 cybrentq Version: 0.1.5 - by Gledis Caushaj (https://github.com/gledi-ai/cybrentq)
 modAB_ct: pymodab version 1.0.5 from PyPI - the previous implementation with ctypes
-modAB_SG: pymodab version 1.1.2 - the latest implementation of the safeguarded modab as native C extension
+modAB_SG: pymodab version 1.1.3 - the latest implementation of the safeguarded modab as native C extension