Backport symlog norm patches to yt-4.0.x

yt/funcs.py

@@ -25,7 +25,7 @@
 import matplotlib
 import numpy as np
 from more_itertools import always_iterable, collapse, first
-from packaging.version import parse as parse_version
+from packaging.version import Version
 from tqdm import tqdm
 
 from yt.units import YTArray, YTQuantity
@@ -1039,7 +1039,7 @@ def matplotlib_style_context(style_name=None, after_reset=False):
         import matplotlib
 
         style_name = {"mathtext.fontset": "cm"}
-        if parse_version(matplotlib.__version__) >= parse_version("3.3.0"):
+        if Version(matplotlib.__version__) >= Version("3.3.0"):
             style_name["mathtext.fallback"] = "cm"
         else:
             style_name["mathtext.fallback_to_cm"] = True

yt/utilities/on_demand_imports.py

@@ -1,6 +1,6 @@
 import sys
 
-from packaging.version import parse as parse_version
+from packaging.version import Version
 
 
 class NotAModule:
@@ -361,7 +361,7 @@ def __init__(self):
         try:
             import h5py
 
-            if parse_version(h5py.__version__) < parse_version("2.4.0"):
+            if Version(h5py.__version__) < Version("2.4.0"):
                 self._err = RuntimeError(
                     "yt requires h5py version 2.4.0 or newer, "
                     "please update h5py with e.g. `python -m pip install -U h5py` "

yt/visualization/base_plot_types.py

@@ -1,8 +1,9 @@
+import warnings
 from io import BytesIO
 
 import matplotlib
 import numpy as np
-from packaging.version import parse as parse_version
+from packaging.version import Version
 
 from yt.funcs import (
     get_brewer_cmap,
@@ -12,7 +13,7 @@
     mylog,
 )
 
-from ._commons import get_canvas, validate_image_name
+from ._commons import MPL_VERSION, get_canvas, validate_image_name
 
 BACKEND_SPECS = {
     "GTK": ["backend_gtk", "FigureCanvasGTK", "FigureManagerGTK"],
@@ -132,9 +133,9 @@ def save(self, name, mpl_kwargs=None, canvas=None):
 
         if mpl_kwargs is None:
             mpl_kwargs = {}
-        if "papertype" not in mpl_kwargs and parse_version(
-            matplotlib.__version__
-        ) < parse_version("3.3.0"):
+        if "papertype" not in mpl_kwargs and Version(matplotlib.__version__) < Version(
+            "3.3.0"
+        ):
             mpl_kwargs["papertype"] = "auto"
 
         name = validate_image_name(name)
@@ -206,20 +207,29 @@ def _init_image(self, data, cbnorm, cblinthresh, cmap, extent, aspect):
             vmin=float(self.zmin) if self.zmin is not None else None,
             vmax=float(self.zmax) if self.zmax is not None else None,
         )
+        zmin = float(self.zmin) if self.zmin is not None else np.nanmin(data)
+        zmax = float(self.zmax) if self.zmax is not None else np.nanmax(data)
+
+        if cbnorm == "symlog":
+            # if cblinthresh is not specified, try to come up with a reasonable default
+            min_abs_val = np.min(np.abs((zmin, zmax)))
+            if cblinthresh is None:
+                cblinthresh = np.nanmin(np.absolute(data)[data != 0])
+            elif zmin * zmax > 0 and cblinthresh < min_abs_val:
+                warnings.warn(
+                    f"Cannot set a symlog norm with linear threshold {cblinthresh} "
+                    f"lower than the minimal absolute data value {min_abs_val} . "
+                    "Switching to log norm."
+                )
+                cbnorm = "log10"
+
         if cbnorm == "log10":
             cbnorm_cls = matplotlib.colors.LogNorm
         elif cbnorm == "linear":
             cbnorm_cls = matplotlib.colors.Normalize
         elif cbnorm == "symlog":
-            # if cblinthresh is not specified, try to come up with a reasonable default
-            vmin = float(np.nanmin(data))
-            vmax = float(np.nanmax(data))
-            if cblinthresh is None:
-                cblinthresh = np.nanmin(np.absolute(data)[data != 0])
-
-            cbnorm_kwargs.update(dict(linthresh=cblinthresh, vmin=vmin, vmax=vmax))
-            MPL_VERSION = parse_version(matplotlib.__version__)
-            if MPL_VERSION >= parse_version("3.2.0"):
+            cbnorm_kwargs.update(dict(linthresh=cblinthresh))
+            if MPL_VERSION >= Version("3.2.0"):
                 # note that this creates an inconsistency between mpl versions
                 # since the default value previous to mpl 3.4.0 is np.e
                 # but it is only exposed since 3.2.0
@@ -275,35 +285,41 @@ def _init_image(self, data, cbnorm, cblinthresh, cmap, extent, aspect):
         if cbnorm == "symlog":
             formatter = matplotlib.ticker.LogFormatterMathtext(linthresh=cblinthresh)
             self.cb = self.figure.colorbar(self.image, self.cax, format=formatter)
-            if np.nanmin(data) >= 0.0:
-                yticks = [np.nanmin(data).v] + list(
+
+            if zmin >= 0.0:
+                yticks = [zmin] + list(
                     10
                     ** np.arange(
                         np.rint(np.log10(cblinthresh)),
-                        np.ceil(np.log10(np.nanmax(data))),
+                        np.ceil(np.log10(zmax)),
                     )
                 )
-            elif np.nanmax(data) <= 0.0:
+            elif zmax <= 0.0:
+                if MPL_VERSION >= Version("3.5.0b"):
+                    offset = 0
+                else:
+                    offset = 1
+
                 yticks = (
                     list(
                         -(
                             10
                             ** np.arange(
-                                np.floor(np.log10(-np.nanmin(data))),
-                                np.rint(np.log10(cblinthresh)) - 1,
+                                np.floor(np.log10(-zmin)),
+                                np.rint(np.log10(cblinthresh)) - offset,
                                 -1,
                             )
                         )
                     )
-                    + [np.nanmax(data).v]
+                    + [zmax]
                 )
             else:
                 yticks = (
                     list(
                         -(
                             10
                             ** np.arange(
-                                np.floor(np.log10(-np.nanmin(data))),
+                                np.floor(np.log10(-zmin)),
                                 np.rint(np.log10(cblinthresh)) - 1,
                                 -1,
                             )
@@ -314,7 +330,7 @@ def _init_image(self, data, cbnorm, cblinthresh, cmap, extent, aspect):
                         10
                         ** np.arange(
                             np.rint(np.log10(cblinthresh)),
-                            np.ceil(np.log10(np.nanmax(data))),
+                            np.ceil(np.log10(zmax)),
                         )
                     )
                 )

yt/visualization/color_maps.py

@@ -1,10 +1,10 @@
 import numpy as np
 from matplotlib import __version__ as mpl_ver, cm as mcm, colors as cc
-from packaging.version import parse as parse_version
+from packaging.version import Version
 
 from . import _colormap_data as _cm
 
-MPL_VERSION = parse_version(mpl_ver)
+MPL_VERSION = Version(mpl_ver)
 del mpl_ver
 
 
@@ -260,7 +260,7 @@ def show_colormaps(subset="all", filename=None):
                 "to be 'all', 'yt_native', or a list of "
                 "valid colormap names."
             ) from e
-    if parse_version("2.0.0") <= MPL_VERSION < parse_version("2.2.0"):
+    if Version("2.0.0") <= MPL_VERSION < Version("2.2.0"):
         # the reason we do this filtering is to avoid spurious warnings in CI when
         # testing against old versions of matplotlib (currently not older than 2.0.x)
         # and we can't easily filter warnings at the level of the relevant test itself

yt/visualization/plot_container.py

@@ -182,8 +182,10 @@ def get_symlog_minorticks(linthresh, vmin, vmax):
         the maximum value in the colorbar
 
     """
-    if vmin > 0 or vmax < 0:
+    if vmin > 0:
         return get_log_minorticks(vmin, vmax)
+    elif vmax < 0 and vmin < 0:
+        return -get_log_minorticks(-vmax, -vmin)
     elif vmin == 0:
         return np.hstack((0, get_log_minorticks(linthresh, vmax)))
     elif vmax == 0:

yt/visualization/plot_window.py

@@ -7,7 +7,7 @@
 import numpy as np
 from more_itertools import always_iterable
 from mpl_toolkits.axes_grid1 import ImageGrid
-from packaging.version import Version, parse as parse_version
+from packaging.version import Version
 from unyt.exceptions import UnitConversionError
 
 from yt._maintenance.deprecation import issue_deprecation_warning
@@ -64,7 +64,7 @@ def zip_equal(*args):
         return zip(*args, strict=True)
 
 
-MPL_VERSION = parse_version(matplotlib.__version__)
+MPL_VERSION = Version(matplotlib.__version__)
 
 # Some magic for dealing with pyparsing being included or not
 # included in matplotlib (not in gentoo, yes in everything else)
@@ -1194,7 +1194,7 @@ def _setup_plots(self):
                     self.plots[f].cax.minorticks_on()
 
                 elif self._field_transform[f] == symlog_transform:
-                    if Version("3.2.0") <= MPL_VERSION < Version("3.5.0"):
+                    if Version("3.2.0") <= MPL_VERSION < Version("3.5.0b"):
                         # no known working method to draw symlog minor ticks
                         # see https://github.com/yt-project/yt/issues/3535
                         pass
@@ -1203,13 +1203,13 @@ def _setup_plots(self):
                             np.log10(self.plots[f].cb.norm.linthresh)
                         )
                         mticks = get_symlog_minorticks(flinthresh, vmin, vmax)
-                        if MPL_VERSION < Version("3.5.0"):
+                        if MPL_VERSION < Version("3.5.0b"):
                             # https://github.com/matplotlib/matplotlib/issues/21258
                             mticks = self.plots[f].image.norm(mticks)
                         self.plots[f].cax.yaxis.set_ticks(mticks, minor=True)
 
                 elif self._field_transform[f] == log_transform:
-                    if MPL_VERSION >= parse_version("3.0.0"):
+                    if MPL_VERSION >= Version("3.0.0"):
                         self.plots[f].cax.minorticks_on()
                         self.plots[f].cax.xaxis.set_visible(False)
                     else:

yt/visualization/profile_plotter.py

@@ -8,7 +8,7 @@
 import numpy as np
 from matplotlib.font_manager import FontProperties
 from more_itertools.more import always_iterable, unzip
-from packaging.version import parse as parse_version
+from packaging.version import Version
 
 from yt.data_objects.profiles import create_profile, sanitize_field_tuple_keys
 from yt.data_objects.static_output import Dataset
@@ -30,7 +30,7 @@
     validate_plot,
 )
 
-MPL_VERSION = parse_version(matplotlib.__version__)
+MPL_VERSION = Version(matplotlib.__version__)
 
 
 def invalidate_profile(f):
@@ -1175,7 +1175,7 @@ def _setup_plots(self):
             if self._cbar_minorticks[f]:
                 if self._field_transform[f] == linear_transform:
                     self.plots[f].cax.minorticks_on()
-                elif MPL_VERSION < parse_version("3.0.0"):
+                elif MPL_VERSION < Version("3.0.0"):
                     # before matplotlib 3 log-scaled colorbars internally used
                     # a linear scale going from zero to one and did not draw
                     # minor ticks. Since we want minor ticks, calculate