Differences

This shows you the differences between two versions of the page.

--- calculate_python_expression [2026/07/19 22:13]
hermann
+++ calculate_python_expression [2026/08/31 17:10] (current)
hermann
@@ Line 1: / Line 1: @@
-====== Calculate R Expression ======
+====== Calculate Python Expression ======
 ===== Description =====
-This is a **[[ego_script#container_functors|container functor]]** that calls R externally and processes the script's outputs as if they were part of Dinamica EGO itself. Like the other calculator functors, data is connected through hook functors placed inside its ''<nowiki>{{ … }}</nowiki>'' block.
+This is a container functor that runs a Python instance with a user-defined expression. Like the other calculator functors, such as [[Calculate Map]] or [[Calculate Lookup Table]], data is connected through hook functors placed inside its block, rather than through regular input ports. See also [[Calculate R Expression]] for the equivalent functor using R.
 ===== Inputs =====
-^ Name ^ Type ^ Description ^
+^ Name  ^ Type  ^ Description  ^
-| Expression | [[ego_script#constants|Code]] | The expression that will run on R. ''Code'' values cannot be written as plain text constants in EGO Script — see [[#writing_the_expression_in_ego_script|Writing the expression in EGO Script]] below. |
+| Expression  | [[Code Type]]  | The expression to run on Python. Written directly as a Code constant using its own raw string syntax -- see [[#writing_the_expression_in_ego_script|Writing the expression in EGO Script]] below.  |
-| Treat Warning As Errors | [[Boolean Value Type]] | Warnings raised by the R script will be treated as errors. |
 ===== Optional Inputs =====
-None.
+^ Name  ^ Type  ^ Description  ^ Default Value  ^
+| Packages  | [[String Type]]  | Packages to install with pip before the expression runs, one per line. Each package can be identified by name or by a filename or URL pointing to the corresponding wheel; a package already installed is skipped. This is an advanced port.  | None  |
 ===== Outputs =====
-^ Name ^ Type ^ Description ^
+^ Name  ^ Type  ^ Description  ^
-| result | [[struct_type|Struct]] | A struct containing the output values generated by the expression. |
+| Result  | [[Struct Type]]  | Struct containing the output values generated by the expression, one entry per key assigned into dinamica.outputs.  |
+===== Group =====
+[[Functor List#Integration | Integration]]
 ===== Notes =====
@@ Line 24: / Line 28: @@
 ==== Expression inputs ====
-Data is passed into the expression through **hook** functors placed inside the container's ''<nowiki>{{ … }}</nowiki>'' block. Hooks can be added using the //Create a hook// button on the functor bar, or by dragging them in individually.
+Data is passed into the expression through hook functors placed inside the container's block -- the same verbose-form hook mechanism used by the calculator functors:
-  * Tables and lookup tables → [[Number Table]] → available in R as ''t1'', ''t2'', …, ''t100''
+  * Tables and lookup tables → [[Number Table]] → available in the expression as dinamica.inputs["t1"], dinamica.inputs["t2"], …, dinamica.inputs["t100"]
-  * Scalar values → [[Number Value]] → available as ''v1'', ''v2'', …, ''v100''
+  * Scalar values → [[Number Value]] → available as dinamica.inputs["v1"], dinamica.inputs["v2"], …, dinamica.inputs["v100"]
-  * Strings → [[Number String]] → available as ''s1'', ''s2'', …, ''s100''
+  * Strings → [[Number String]] → available as dinamica.inputs["s1"], dinamica.inputs["s2"], …, dinamica.inputs["s100"]
-Maps cannot be connected — this functor has no cell context. There is no shorthand notation; see [[calculate_functors|Calculate Functors — Complete Operator Documentation]] for the general hook mechanism and syntax.
+Maps cannot be connected -- this functor has no cell context. There is no shorthand notation. See [[Calculate Functors|Calculate Functors -- Complete Operator Documentation]] for the general hook mechanism and syntax.
-=== Tables and lookup tables ===
+Every table or lookup table arriving through dinamica.inputs is represented in Python as a list of lists: the first inner list contains the column names (the header row) and every subsequent inner list is a row of data. The header carries plain column names only -- no asterisk marking key columns, no type annotation -- so which columns were keys in the original table is information Python doesn't receive and can't recover from the input alone. Code that needs that distinction has to be told it separately, for instance by also connecting the key count as its own [[Number Value]] hook.
-Lookup tables and tables require extra care, since each is transferred to R using a different representation.
+==== Expression outputs ====
-A **lookup table** is transferred as a list with two columns, ''Key'' and ''Value''. Each column is accessed with the ''$'' operator:
+Values are returned to Dinamica by assigning them into dinamica.outputs, keyed by the desired output name. Every assigned value becomes an entry in the output Result struct:
-<code rsplus>
+<code python>
-# Access the third key and its corresponding value in lookup table t1
+# Scalar values are assigned directly
-patchId <- t1$Key[3];
+dinamica.outputs["patchCount"] = 42
-patchArea <- t1$Value[3];
+dinamica.outputs["totalArea"] = 1530.5
 </code>
-A **table** is transferred as a [[https://en.wikibooks.org/wiki/R_Programming/Working_with_data_frames|DataFrame]], with each column likewise accessed using ''$''. Two conventions apply to tables in either direction:
+Tables and lookup tables cannot be assigned directly -- they must first be converted using the utilities described below. If a table arriving from an input already carries asterisk markers on its key column names, it can be assigned directly to an output without conversion.
-  * **Key columns** are marked by an asterisk (''*'') appended to their column name — this is the same convention used throughout Dinamica EGO's table representation (see [[calculate_functors#connecting_data_inputs|Connecting Data Inputs]]).
+==== Retrieving outputs ====
-  * R automatically converts string columns to **Factors** inside a ''data.frame'', but Dinamica requires plain **Character Vectors**. Always build tables with ''stringsAsFactors = FALSE'' to prevent this conversion.
-For further detail on the underlying table representation, see [[external_communication#table|External Communication]].
+Calculate Python Expression returns a single [[Struct Type]] value (via its Result output port) containing every entry assigned to dinamica.outputs. To retrieve individual values from that struct, use the following functors from the Integration group:
-==== Expression outputs ====
+^ Functor  ^ Retrieves  ^
+| [[Extract Struct Number]]  | A numeric value (int or float assigned to dinamica.outputs)  |
+| [[Extract Struct String]]  | A string value  |
+| [[Extract Struct Table]]  | A table produced by dinamica.prepareTable()  |
+| [[Extract Struct Lookup Table]]  | A lookup table produced by dinamica.prepareLookupTable()  |
+| [[Extract Struct Tuple]]  | A tuple value  |
-Values are returned to Dinamica by calling one of the following functions from the R script. Every call requires an identifier as its first parameter — the name Dinamica uses to place the value into the output struct:
+Each functor takes two inputs: the Struct returned by Calculate Python Expression, and the name of the entry to extract as a string constant. For example, to retrieve a numeric output named patchCount and a table output named filteredPatches:
-^ Function ^ Output type ^ Example ^
+<code>
-| ''outputDouble()'' | Real | ''outputDouble("myDouble", 3.14)'' |
+result := CalculatePythonExpression $"(
-| ''outputNumberVector()'' | Tuple | ''outputNumberVector("myTuple", c(1:10))'' |
+myTable = [["PatchId*", "Area"], [1, 12.4], [2, 8.7]]
-| ''outputString()'' | String | ''outputString("myString", "This is a string")'' |
+dinamica.outputs['patchCount'] = 42
-| ''outputLookupTable()'' | Lookup Table | ''outputLookupTable("myLUT", c(1:10), c(1:10) * 10)'' |
+dinamica.outputs['filteredPatches'] = dinamica.prepareTable(myTable, 1)
-| ''outputTable()'' | Table | ''outputTable("myTable", data.frame(State = c("Massachusetts", "Massachusetts"), City = c("Boston", "Chelsea"), Population = c(667137, 39398), stringsAsFactors = FALSE), 2)'' |
+)" {{ }};
+patchCount      := ExtractStructNumber result "patchCount";
+filteredPatches := ExtractStructTable  result "filteredPatches";
+</code>
-Values passed to these functions can be constructed inline, as in the examples above, or supplied as variables.
+==== Utilities ====
-  * Any numeric value can be passed to ''outputDouble()''.
+=== Installing packages ===
-  * Collections of numbers are valid number vectors for ''outputNumberVector()''.
-  * A [[Lookup Table Type|lookup table]] requires two number vectors of equal length — one for the keys, one for the values.
-  * ''outputTable()'' requires a table built with the [[https://www.r-tutor.com/r-introduction/data-frame|data.frame]] function, using ''stringsAsFactors = FALSE'' as described above. Its optional second parameter (default ''1'') sets how many leading columns, from the left, are key columns.
-==== Retrieving outputs ====
+Packages can be installed in two ways: calling dinamica.package(...) from within the expression, or listing package names on the Packages input port. The two differ in timing: dinamica.package(...) runs during the expression, so it can be called conditionally; the Packages port always runs before the expression executes.
-''CalculateRExpression'' returns a single [[struct_type|Struct]] value (via its ''result'' output port) containing every value passed to an ''output*()'' function. To retrieve individual values from that struct, use the corresponding functor from the //Integration// group:
+**dinamica.package()**
-^ Functor ^ Retrieves ^
+dinamica.package(packageName, installPath=None, loadPath=None) installs (if needed, via pip) and imports the requested module. Because it runs together with the script, it may fail if a package is already loaded in an incompatible version.
-| [[Extract Struct Number]] | A value passed to ''outputDouble()'' |
-| [[Extract Struct Tuple]] | A value passed to ''outputNumberVector()'' |
-| [[Extract Struct String]] | A value passed to ''outputString()'' |
-| [[Extract Struct Lookup Table]] | A value passed to ''outputLookupTable()'' |
-| [[Extract Struct Table]] | A value passed to ''outputTable()'' |
-Each functor takes two inputs: the ''Struct'' returned by ''CalculateRExpression'', and the name of the entry to extract as a string constant.
+^ Parameter  ^ Type  ^ Default  ^ Description  ^
+| packageName  | str  | --  | Identifies the package. Used as the default for both installPath and loadPath when those are omitted.  |
+| installPath  | str  | packageName  | What is passed to pip install. Can be a plain name, a version-pinned requirement, a wheel filename or URL, a git+ URL, or a name followed by extra pip flags.  |
+| loadPath  | str  | packageName  | The name used to import the module in Python.  |
-==== Installing packages ====
+Simply install and load numpy:
-Packages are installed by calling ''dinamicaPackage("packageName")'' from within the expression — one call per package. Unlike [[calculate_python_expression|Calculate Python Expression]], there is no separate input port for listing packages; ''dinamicaPackage()'' is the only mechanism available.
+<code python>
+dinamica.package("numpy")
+</code>
-''dinamicaPackage("packageName")'' also acts as R's ''library()'' call: when the package name matches the name of the module to load, calling it both installs the package (if not already present) and loads it, in a single call.
+Specify a version:
-<code rsplus>
+<code python>
-dinamicaPackage("moments");
+dinamica.package("numpy", "numpy==1.19.5")
 </code>
-==== Setup ====
+Install a package whose importable name differs from its pip name:
-There are two ways to run R scripts from ''CalculateRExpression'' — though only the local installation is available on Linux.
+<code python>
+dinamica.package("segment_anything_py", "segment_anything_py", "segment_anything")
+</code>
-=== Dinamica EGO Enhancement Plugin ===
+Use arbitrary pip parameters, such as installing from a remote wheel with a custom index:
-Windows only. Download and install the [[plugins_4|Dinamica EGO Enhancement Plugin]]. It contains everything needed to run R scripts inside Dinamica EGO, with no further configuration.
+<code python>
+dinamica.package("segment_anything_py", "https://files.pythonhosted.org/packages/43/2f/dabe75d90a7eb54a0a609a0fc5c36d1933256319beaea5d6b2f176e213a2/segment_anything_py-1.0-py3-none-any.whl --index-url https://download.pytorch.org/whl/cu118", "segment_anything")
+</code>
-=== Local R installation ===
+Chain several installs:
-On Linux, this is the only option — Dinamica EGO always uses the R installation already present on the system. On Windows, it can be used as an alternative to the plugin. Either way, it requires:
+<code python>
+dinamica.package("cython")
+dinamica.package("numpy")
+dinamica.package("pycocotools", "git+https://github.com/philferriere/cocoapi.git#egg=pycocotools&subdirectory=PythonAPI")
+</code>
-  * R installed on the machine, with the ''Rscript'' executable (''Rscript.exe'' on Windows) present in its ''bin'' sub-folder.
+**The Packages port**
-  * The [[external_communication|Dinamica package]] for R installed and at its latest version.
-On Windows, this alternative installation is selected in the Dinamica EGO GUI by going to //Tools// → //Options// → //Integration// tab and enabling //Use alternative R installation for Calculate R Expression//.
+The Packages input port takes one package identifier per line, in any form accepted by pip. Every listed package is installed before the expression runs. Unlike dinamica.package(), the port only installs -- it does not import anything. Importing is left entirely to the expression, using standard Python import syntax and each package's actual importable name, which may differ from the name given to pip.
-==== Examples ====
+The Packages port:
-The following examples use a consistent set of inputs:
+<code>
+numpy
+requests==2.31.0
+torchvision==0.19.1 --index-url https://download.pytorch.org/whl/cu121
+</code>
-  * ''t1'' — a lookup table of land cover patches, mapping ''Key'' (patch identifier) to ''Value'' (patch area)
+The expression:
-  * ''v1'' — a scalar minimum area threshold
-  * ''s1'' — a string giving the name to use for the threshold flag column
+<code python>
+import numpy as np
+from requests import Session
+from torchvision import transforms
+</code>
+=== dinamica.prepareTable() ===
+Converts a list of lists into a table ready to be assigned to an output. The first inner list must be the header row. As with any table, each column's type is inferred from its own values: a column of int/float values produces a Real column, a column of str values produces a String column.
+^ Parameter  ^ Type  ^ Default  ^ Description  ^
+| inputTable  | list of lists  | --  | The table data. First inner list is the header row; subsequent inner lists are data rows.  |
+| numKeys  | int  | --  | Number of key columns. Key column names will have an asterisk appended in the output.  |
+=== dinamica.prepareLookupTable() ===
+Converts a list of lists into a lookup table ready to be assigned to an output. The first inner list must be the header row. Lookup tables are always Real-typed on both key and value sides -- there is no String option -- so every value in lut must be numeric.
+^ Parameter  ^ Type  ^ Default  ^ Description  ^
+| lut  | list of lists  | --  | The lookup table data. First inner list is the header row; subsequent inner lists are data rows.  |
+=== dinamica.toTable() ===
+Converts several Python data shapes into a valid Dinamica table for output. Column types are inferred the same way as dinamica.prepareTable() -- except for a flat list, which always produces a Real-typed lookup table with sequential keys, matching the Real-only rule for lookup tables.
+^ Parameter  ^ Type  ^ Default  ^ Description  ^
+| inputTable  | list of lists; dict of lists; list of tuples; flat list; pandas.DataFrame; numpy.array  | --  | The data to convert. A flat list produces a lookup table with sequential keys. A numpy.array must have its header as the first row.  |
+| numKeys  | int  | --  | Number of key columns. Key column names will have an asterisk appended in the output. Ignored for a flat list, which always produces a lookup table with sequential keys.  |
+==== Examples ====
-Compute the mean and total area across all patches, and report progress to the Message Log:
+The following examples use a consistent set of inputs: t1 is a table of land cover patches with columns PatchId*, Area, and CategoryId; t2 is a lookup table mapping category identifiers to category names; and v1 is a scalar minimum area threshold.
-<code rsplus>
+Install and import numpy, then inspect all inputs passed in by Dinamica:
-patchMean <- mean(t1$Value);
-patchTotal <- sum(t1$Value);
-print(paste("Processed", length(t1$Value), "patches"));
-outputDouble("meanArea", patchMean);
+<code python>
-outputDouble("totalArea", patchTotal);
+dinamica.package("numpy")
+print(dinamica.inputs)
 </code>
-> **Note:** The ''print()'' call is visible in Dinamica EGO's Message Log, shown as a Result-level message. Log levels are ordered Unconditional, Error, Warning, Result, Info, Info2, Debug, Debug2 — messages printed from R are only shown when the Message Log level is set to Result or a more verbose level; at Unconditional, Error, or Warning they are suppressed.
+The print() call is visible in Dinamica EGO's Message Log, shown as a Result-level message. Log levels are ordered Unconditional, Error, Warning, Result, Info, Info2, Debug, Debug2 -- messages printed from Python are only shown when the Message Log level is set to Result or a more verbose level; at Unconditional, Error, or Warning they are suppressed.
-Filter the lookup table down to patches whose area meets the threshold, and return the filtered lookup table:
+Print the rows of both connected tables to verify their contents:
-<code rsplus>
+<code python>
-aboveThreshold <- t1$Value >= v1;
+for row in dinamica.inputs["t1"]:
-filteredKeys <- t1$Key[aboveThreshold];
+    print(row)
-filteredValues <- t1$Value[aboveThreshold];
-outputLookupTable("filteredPatches", filteredKeys, filteredValues);
+for row in dinamica.inputs["t2"]:
+    print(row)
 </code>
-Build and return a table with one row per patch, including a column that flags whether each patch meets the threshold. The column's name is taken from the passed string ''s1'' rather than being hard-coded:
+Count the patches whose area meets the minimum threshold and sum their total area:
-<code rsplus>
+<code python>
-patchFlags <- data.frame(
+patchCount = 0
-    PatchId = t1$Key,
+totalArea = 0.0
-    Area = t1$Value,
+for row in dinamica.inputs["t1"][1:]:
-    MeetsThreshold = t1$Value >= v1,
+    if row[1] >= dinamica.inputs["v1"]:
-    stringsAsFactors = FALSE
+        patchCount += 1
-);
+        totalArea += row[1]
-names(patchFlags)[3] <- s1;
-outputTable("patchSummary", patchFlags, 1);
+dinamica.outputs["patchCount"] = patchCount
+dinamica.outputs["totalArea"] = totalArea
 </code>
-Install the ''moments'' package and use it to compute the skewness of the patch area distribution — a statistic not available in base R — then flag patches whose area is a statistical outlier:
+Return a filtered table containing only the patches above the threshold:
-<code rsplus>
+<code python>
-dinamicaPackage("moments");
+filtered = [dinamica.inputs["t1"][0]]
+for row in dinamica.inputs["t1"][1:]:
+    if row[1] >= dinamica.inputs["v1"]:
+        filtered.append(row)
-patchMean <- mean(t1$Value);
+dinamica.outputs["filteredPatches"] = dinamica.prepareTable(filtered, 1)
-patchStdDev <- sd(t1$Value);
+</code>
-patchSkewness <- skewness(t1$Value);
-isOutlier <- abs(t1$Value - patchMean) > 2 * patchStdDev;
+Compute the total area per category and return it as a lookup table:
-outlierPatches <- data.frame(
-    PatchId = t1$Key[isOutlier],
-    Area = t1$Value[isOutlier],
-    stringsAsFactors = FALSE
-);
-outputDouble("meanArea", patchMean);
+<code python>
-outputDouble("stdDevArea", patchStdDev);
+areaSums = {}
-outputDouble("skewnessArea", patchSkewness);
+for row in dinamica.inputs["t1"][1:]:
-outputTable("outlierPatches", outlierPatches, 1);
+    categoryId = row[2]
+    areaSums[categoryId] = areaSums.get(categoryId, 0.0) + row[1]
+lut = [["CategoryId*", "TotalArea"]]
+for catId, total in areaSums.items():
+    lut.append([catId, total])
+dinamica.outputs["areaByCategory"] = dinamica.prepareLookupTable(lut)
 </code>
 ==== Writing the expression in EGO Script ====
-Like [[calculate_python_expression|Calculate Python Expression]], the ''Expression'' input is type [[ego_script#constants|Code]], which is represented in the underlying script using base64 encoding — impractical to write or edit directly as a text constant. Instead, connect a ''String'' carrier functor containing the R expression text to the ''Expression'' port; its output is accepted wherever a ''Code'' value is expected. Since only this one output is needed, the carrier can be [[ego_script#inline_syntax|inlined]] directly into the call.
+Expression can be filled in directly as a text constant, using [[Code Type]]'s own raw string syntax -- the same ''$"<delimiter>( raw_characters )<delimiter>"'' form used by String constants. This is also the form the Dinamica EGO GUI's dedicated code editor generates when it writes the Expression port's value, so a hand-written script and one produced by the GUI take the same shape. See [[Code Type]] for the full grammar, including the base64 alternative form.
-This limitation is specific to hand-written EGO Script. In the Dinamica EGO GUI, the ''Expression'' port has a dedicated code editor that edits the ''Code'' value directly — the ''String'' carrier workaround is only necessary when writing or editing the ''.ego'' file as text.
+The following counts the patches in a land cover areas table whose area meets a minimum threshold:
 <code>
-result := CalculateRExpression (String $"(
+result := CalculatePythonExpression $"(
-aboveThreshold <- t1$Value >= v1;
+total = 0
-outputLookupTable("filteredPatches", t1$Key[aboveThreshold], t1$Value[aboveThreshold]);
+for row in dinamica.inputs['t1'][1:]:
-)") .no {{
+    if row[1] >= dinamica.inputs['v1']:
+        total += 1
+dinamica.outputs['patchCount'] = total
+)" {{
     NumberTable landCoverAreas 1;
     NumberValue minimumArea    1;
 }};
-filteredPatches := ExtractStructLookupTable result "filteredPatches";
+patchCount := ExtractStructNumber result "patchCount";
 </code>
-===== Group =====
+Here landCoverAreas is bound to t1 and minimumArea to v1, each through a [[Number Table]] or [[Number Value]] hook. Calculate Python Expression returns a Struct containing all values assigned to dinamica.outputs; the ExtractStructNumber functor then pulls the patchCount entry out of that struct by name. See "Retrieving outputs" above for the full list of extraction functors.
-[[Functor List#Integration|Integration]]
+A more involved example: installing numpy to compute area statistics for a land cover patches table and flag outlier patches, then retrieving both the scalar statistics and the resulting table:
-===== Internal Name =====
+<code>
+result := CalculatePythonExpression $"(
+dinamica.package('numpy')
-CalculateRExpression
+areas = numpy.array([row[1] for row in dinamica.inputs['t1'][1:]])
-===== See Also =====
+meanArea = float(numpy.mean(areas))
+stdArea = float(numpy.std(areas))
-  * [[calculate_python_expression|Calculate Python Expression]]
+isOutlier = numpy.abs(areas - meanArea) > 2 * stdArea
-  * [[External Communication]]
-  * [[ego_script#container_functors|EGO Script — Container Functors]]
-  * [[ego_script#verbose_form|EGO Script — Verbose Form (Hooks)]]
-  * [[calculate_functors|Calculate Functors — Complete Operator Documentation]]
+dinamica.outputs['meanArea'] = meanArea
+dinamica.outputs['stdArea'] = stdArea
+header = dinamica.inputs['t1'][0]
+outlierRows = [row for row, flagged in zip(dinamica.inputs['t1'][1:], isOutlier) if flagged]
+outlierTable = [header] + outlierRows
+dinamica.outputs['outlierPatches'] = dinamica.prepareTable(outlierTable, 1)
+)" {{
+    NumberTable landCoverPatches 1;
+}};
+meanArea       := ExtractStructNumber result "meanArea";
+stdArea        := ExtractStructNumber result "stdArea";
+outlierPatches := ExtractStructTable  result "outlierPatches";
+</code>
+Here landCoverPatches is bound to t1 through a single [[Number Table]] hook.
+===== Internal Name =====
+CalculatePythonExpression