[MatL] Multi-phase/-component library. #2303

endJunction · 2018-12-21T11:22:34Z

This is part of #1955 which is used in multi component transport (#2304). Mainly infrastructure, since only constant properties were used so far in the ComponentTransportProcess.

TomFischer · 2019-01-08T13:08:01Z

The copyright has to be updated.

TomFischer · 2019-01-08T13:09:17Z

MaterialLib/MPL/Enums.h

+/// determine the size of the VariableArray. If the variable of your choice is
+/// missing, simply add it somewhere at the list, but above the last entry
+/// 'numberOfVariables'
+enum Variables : std::size_t


Why not 'enum class'?

The implicit conversion to int (I changed the underlying type) is quite handy.

TomFischer · 2019-01-08T13:10:56Z

MaterialLib/MPL/Enums.h

+#pragma once
+
+#include <boost/algorithm/string/predicate.hpp>
+#include <boost/variant.hpp>


Can we already use std::variant?

Unfortunately not until gcc-7.1. And also the macOS's clang in version 10 does not support it fully yet. But (fasten your belts!!!) the latest visual studio would work!

Also tried new syntax for "structured binding declarations" instead of std::tie

auto [a, b, c] = function_returning_a_tuple_of_tree();

working only since gcc-7

TomFischer · 2019-01-08T13:11:52Z

MaterialLib/MPL/Enums.h

+/// in alphabetical order (of course, except for the last entry
+/// 'number_of_property_enums'). Please note that any of these entries must also
+/// appear in below convert functions.
+enum PropertyType : std::size_t


Why not enum class?

MaterialLib/MPL/Enums.h

TomFischer · 2019-01-08T13:20:13Z

MaterialLib/MPL/MaterialSpatialDistributionMap.h

+class MaterialSpatialDistributionMap
+{
+private:
+    std::map<int, std::unique_ptr<Medium>> const& _media;


As far as I understand the domain of the map is [0, #number of media]. Maybe a vector can be used here?

Note to myself; compare with error/warning message code.

@TomFischer the media ids follow numbering from MaterialIDs and can be rather arbitrary, especially negative. We might have a look on it later, if it looks like a bottleneck.

MaterialLib/MPL/Medium.h

wenqing · 2019-01-08T13:45:39Z

Applications/ApplicationsLib/ProjectData.cpp

@@ -26,6 +26,7 @@
 #include "BaseLib/FileTools.h"

 #include "GeoLib/GEOObjects.h"
+#include "MaterialLib/MPL/mpMedium.h"


Not for this PR: In future I would like to replace MPL with MaterialPropertyLib. Also, it is better to replace mpMedium.h` with some meaningful name.

Applications/ApplicationsLib/ProjectData.cpp

wenqing · 2019-01-08T14:00:13Z

Applications/ApplicationsLib/ProjectData.cpp

+    if (_media.rbegin()->first != static_cast<int>(_media.size()) - 1)
+           OGS_FATAL(
+               "The ids in the porous media definitions in the project file have "
+               "to be sequential, starting with the id zero.");


the id zero --> the id of zero.

I dropped the whole if-condition. The requirement of media to be numbered from 0 to n sequentially, does not match with the freedom of the MaterialIDs, which can be any int, especially negative.

wenqing · 2019-01-08T14:19:19Z

MaterialLib/MPL/Component.cpp

+{
+    // If a name is given, it must conform with one of the derived component
+    // names in the following list:
+    if (boost::iequals(component_name, "water"))


Why not if (component_name == "water")?

Norbert's argument is that it is difficult to explain people, why "Water" is not good enough for the specification of water. Same goes with "CO2" etc.

wenqing · 2019-01-08T14:21:06Z

MaterialLib/MPL/Component.h

+ */
+#pragma once
+
+#include "BaseLib/ConfigTree.h"


can use forward declaration.

Whole creation of media, phases, components, and properties is now moved into corresponding CreateX.h/cpp files.

wenqing · 2019-01-08T15:07:19Z

MaterialLib/MPL/Properties/Constant.cpp

@@ -0,0 +1,25 @@
+/**
+ * \file


remove it or leave a blank line after it.

wenqing · 2019-01-08T15:08:15Z

MaterialLib/MPL/Properties/Properties.h

+ */
+#pragma once
+
+#include "Constant.h"


This file only contains one include.

Same as for components, the full implementation has many, and the properties include is specified once in the process.
Other properties (and also components) which are not tested in the follow-up PR, are excluded.

wenqing · 2019-01-08T15:09:03Z

MaterialLib/MPL/Property.h

+#include <boost/variant.hpp>
+#include <string>
+
+#include "BaseLib/ConfigTree.h"


Forward declaration.

wenqing · 2019-01-08T15:10:11Z

Tests/MaterialLib/TestMPL.h

@@ -0,0 +1,21 @@
+/**
+ * \file


Leave a blank line after it.

Tests/MaterialLib/TestMPLParseMaterial.cpp

MaterialLib/MPL/Component.cpp

TomFischer · 2019-01-09T10:36:00Z

MaterialLib/MPL/Property.h

+};
+
+using PropertyDataType =
+    boost::variant<double, Pair, Vector, SymmTensor, Tensor, std::string>;


I think std::variant can be used.

wenqing

Looks good! Only one idea about MaterialPropertyLib::PropertyType and its converters.

wenqing · 2019-01-14T12:47:15Z

MaterialLib/MPL/CreateProperty.cpp

+        std::stringstream issValue(sValue);
+
+        std::vector<double> values;
+        double dummy(0);


dummy cann't be const since it is used to read from the stringstream issValues. So maybe 'dummy' isn't so dummy ;-)

Didn't see the reading of std::vector<double> from the config. ConfigTree supports getConfigParameter<std::vector<T>>(...). The whole section is replaced by single statement.

wenqing · 2019-01-14T16:08:17Z

MaterialLib/MPL/PropertyType.h

+/// I suggest that the syntax of the properties in the input-files (i.e. the
+/// strings) have to be identical to the syntax of the entries in the
+/// enumerator.
+inline PropertyType convertStringToProperty(std::string const& inString)


If we want to the elements of PropertyType, we have to to change strings in convertStringToProperty and property_enum_to_string as well.

~~It might be good to~~
~~1) create a enum to std::string convertor in BaseLib (see https://stackoverflow.com/questions/28828957/enum-to-string-in-modern-c11-c14-c17-and-future-c20).~~
~~2) use the enum to std::string convertor to initialize std::array<std::string, PropertyType::number_of_properties> property_enum_to_string~~
~~3) create a std::map<std::string, PropertyType> by using of std::string convertor to replace~~ ~convertStringToProperty ~

~~With that, we only need to change the elements of PropertyType if such change is needed.~~

The current implementation is also OK because there might be less change in PropertyType.

1) is not good. One still has to use the enum name to get its string type. It seems that each name of enum element must appears at least two times in the source code. Therefore, please go forward with the present implementation.

TomFischer · 2019-01-15T13:38:40Z

MaterialLib/MPL/Component.h

+              std::unique_ptr<PropertyArray>&& properties);
+    virtual ~Component() = default;
+
+    /// A get-function for retrieving a cartain property.


cartain -> certain

TomFischer · 2019-01-15T13:41:09Z

MaterialLib/MPL/Component.h

+        return property(p).template value<T>();
+    }
+
+    template <typename T>


Is my understanding correct: the variable_array is used to pass information for property value calculation?

TomFischer · 2019-01-15T14:01:50Z

MaterialLib/MPL/CreatePhase.cpp

+        ERR("Phase type should be one of:");
+        for (auto const type : allowed_phase_types)
+        {
+            ERR(type.c_str());


Is this valid? I thought it must be ERR("%s", type.c_str()).

It's just a char array, with substitution or without.

TomFischer · 2019-01-16T06:40:06Z

MaterialLib/MPL/Property.h

+    return p.value().which();
+}
+
+inline void overwriteExistingProperties(PropertyArray& properties,


Since in line 111 new_properties entries are moved, the new_properties aren't valid anymore, isn't it? Maybe the name of the function is a bit misleading, but I haven't a good suggestion.

Good point. Need a name.

TomFischer · 2019-01-16T06:47:51Z

MaterialLib/MPL/VariableType.h

+
+/// Simple symmetric tensor data type for holding
+/// xx, yy, zz, xy, xz, yz tensor components.
+using SymmTensor = std::array<double, 9>;


The documentation says there are 6 entries, the array has size 9. Is this correct?

Good catch! fixing.

TomFischer · 2019-01-16T06:52:13Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+}
+
+/// This method creates an XML-snippet of a sigle property.
+std::string makeProperty(std::size_t ind, std::size_t index,


Why not ... makeProperty(std::size_t const ind, std::size_t const index, std::string const& property)?

TomFischer · 2019-01-16T06:55:41Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+{
+    std::stringstream sProperty;
+
+    if (property == "")


Maybe property.empty() would be more readable.

TomFischer · 2019-01-16T06:58:20Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+    // What follows is a C-style string-to-double-conversion...
+    // seems so easy compared to c++-standard; Feel free to
+    // change it to latest standard..
+    if (double value = atof(property.c_str()))


std::atof please.

TomFischer · 2019-01-16T07:00:17Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+    return sComponents.str();
+}
+
+/// This method creates an XML-snippet of a sigle material


sigle -> single

TomFischer · 2019-01-16T07:01:17Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+
+/// This method creates an XML-snippet of the material
+/// phases.
+std::string makePhases(std::vector<Phase> phases)


std::vector<Phase> const& phases?

TomFischer · 2019-01-16T07:02:09Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+/// This method creates the entire XML-tree structure from the string-
+/// based medium specification object. I know, indentation is not
+/// necessary for this, but my OCD kicked in...
+std::string makeMedium(Medium m)


Medium const& m

TomFischer · 2019-01-16T07:03:50Z

Tests/MaterialLib/TestMPLParseMaterial.cpp

+              std::vector<std::string>* exp)
+{
+    obs->push_back(observed_name);
+    if (expectation == "")


IMHO expectation.empty() would be more readable.

TomFischer

Looks good. ⏩

TomFischer · 2019-01-16T07:20:01Z

MaterialLib/MPL/Phase.h

+private:
+    std::vector<std::unique_ptr<Component>> const _components;
+
+    /// Here, all for all properties listed in the Properties enumerator are


~~all for~~

endJunction · 2019-01-16T10:21:33Z

@TomFischer I polished the TestMPLParseMaterial.cpp with all the const, &, typos, etc. being fixed now.

endJunction · 2019-01-16T10:22:52Z

Two open points remain: enum <-> string, and name for overwriteProperties.

A lot of changes are renames for consistency with OGS. Splitting files, and rename them (dropping mp, p, c prefixes). Extraction of the create* functions lead to single step initialization of the Medium, Phase, Component, and Property objects. Formatting and clang-tidy. Improve const-correctness. Remove unneeded Undefined property: The nullptr is probably sufficient in the PropertyArray, which is an array of pointers, for the property to be undefined. Unify value/dValue/d2Value access. [App] Remove requirement for media ids sequence. Reorder public, protected, private sections. Add missing ogs_file_param documentation lines in the code.

endJunction · 2019-01-17T08:44:10Z

When there are no further comments, and the builds pass, this can be merged.
The enum <-> string conversion needs additional discussions first.

wenqing · 2019-01-17T09:24:05Z

⏩

ogsbot · 2020-06-19T12:56:48Z

OpenGeoSys development has been moved to GitLab.

See this pull request on GitLab.

renchao-lu mentioned this pull request Dec 21, 2018

[PL] Multi-component transport process #2304

Merged

endJunction force-pushed the gpml branch from 9aa48a5 to 3064216 Compare December 21, 2018 15:40

endJunction force-pushed the gpml branch 3 times, most recently from 9a47f59 to b1c7d3b Compare January 4, 2019 18:54

TomFischer reviewed Jan 8, 2019

View reviewed changes

MaterialLib/MPL/Enums.h Outdated Show resolved Hide resolved

TomFischer reviewed Jan 8, 2019

View reviewed changes

MaterialLib/MPL/Medium.h Outdated Show resolved Hide resolved

wenqing reviewed Jan 8, 2019

View reviewed changes

MaterialLib/MPL/Component.cpp Outdated Show resolved Hide resolved

TomFischer reviewed Jan 9, 2019

View reviewed changes

endJunction force-pushed the gpml branch 5 times, most recently from b878684 to 90364f4 Compare January 10, 2019 18:57

wenqing approved these changes Jan 14, 2019

View reviewed changes

TomFischer reviewed Jan 15, 2019

View reviewed changes

TomFischer reviewed Jan 16, 2019

View reviewed changes

TomFischer approved these changes Jan 16, 2019

View reviewed changes

TomFischer reviewed Jan 16, 2019

View reviewed changes

endJunction and others added 4 commits January 17, 2019 09:26

[BL] Add empty() to ConfigTree's Range class.

ebe4ae0

[MatL] Multi-phase/-component library.

4f5e4c3

[Doc] Documentation of <media> prj file tree.

c5442ac

endJunction force-pushed the gpml branch from 6059565 to c5442ac Compare January 17, 2019 08:40

TomFischer merged commit 8972d34 into ufz:master Jan 17, 2019

endJunction deleted the gpml branch January 18, 2019 13:02

endJunction mentioned this pull request Apr 15, 2019

Gpml #1955

Closed

[MatL] Multi-phase/-component library. #2303

[MatL] Multi-phase/-component library. #2303

Conversation

endJunction commented Dec 21, 2018 • edited Loading

TomFischer commented Jan 8, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

wenqing left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction Jan 16, 2019 • edited Loading

Choose a reason for hiding this comment

wenqing Jan 14, 2019 • edited Loading

Choose a reason for hiding this comment

wenqing Jan 15, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TomFischer left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

endJunction commented Jan 16, 2019

endJunction commented Jan 16, 2019

endJunction commented Jan 17, 2019

wenqing commented Jan 17, 2019

ogsbot commented Jun 19, 2020

endJunction commented Dec 21, 2018 •

edited

Loading

endJunction Jan 16, 2019 •

edited

Loading

wenqing Jan 14, 2019 •

edited

Loading

wenqing Jan 15, 2019 •

edited

Loading