Initial Commit of AEO-Public-Release

Author: WhitlingerEIA

.gitattributes

@@ -0,0 +1,15 @@
+# Denote all files that are truly binary and should not be modified.
+# -text means not text aka binary
+# to track in kornshell cd to repo directory and type < git lfs track *xls
+*.dll filter=lfs diff=lfs merge=lfs -text
+*.dsn filter=lfs diff=lfs merge=lfs -text
+*.lib filter=lfs diff=lfs merge=lfs -text
+*.exe filter=lfs diff=lfs merge=lfs -text
+*.xlsx filter=lfs diff=lfs merge=lfs -text
+*.xls filter=lfs diff=lfs merge=lfs -text
+*.wk1 filter=lfs diff=lfs merge=lfs -text
+*.xml filter=lfs diff=lfs merge=lfs -text
+*.gdx filter=lfs diff=lfs merge=lfs -text
+*.mdb filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.daf filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,56 @@
+objects/
+*.obj
+*.mod
+*.unf
+*.pdb
+scripts/backups
+scripts/backup_12_4
+source/backups
+p1/
+p2/
+p3/
+source/build/
+ERROR
+
+
+models/coal/log/*
+!models/coal/log/leave_blank.txt
+models/coal/toAIMMS/*
+!models/coal/toAIMMS/leave_blank.txt
+models/coal/fromAIMMS/*
+!models/coal/fromAIMMS/leave_blank.txt
+
+models/ngas/ngsteo.txt
+models/ngas/ng_runval.txt
+models/ngas/spot_prc_diff.txt
+models/ngas/NGATTENTION.txt
+models/ngas/NGPRICES.txt
+models/ngas/data/NGSTEOFactors.txt
+models/ngas/mainproject/Main.DeveloperState
+models/ngas/mainproject/PageManager.xml
+models/ngas/mainproject/TemplateManager.xml
+models/ngas/mainproject/Pages/*
+models/ngas/mainproject/Settings/*
+models/ngas/mainproject/Templates/*
+models/ngas/mainproject/Tools/*
+models/ngas/mainproject/User*Files/*
+models/ngas/log/*
+!models/ngas/log/leave_blank.txt
+models/ngas/toAIMMS/*
+!models/ngas/toAIMMS/leave_blank.txt
+models/ngas/fromAIMMS/*
+!models/ngas/fromAIMMS/leave_blank.txt
+
+models/rest/fromAIMMS/RETURNCODE.txt
+models/rest/log/*
+!models/rest/log/leave_blank.txt
+models/rest/monitor*.txt
+
+models/efd/log/*
+!models/efd/log/leave_blank.txt
+
+models/ecp/log/*
+!models/ecp/log/leave_blank.txt
+models/ecp/DLL_*.*
+models/ecp/OutToNEMS_*.txt
+models/ecp/monitor*.txt

analyze/!AIX.HLP

@@ -0,0 +1,11 @@
+    Function: Execute AIX command
+    Syntax:   !AIX command_string
+
+    Consult AIX manual.
+
+    Examples:
+      !AIX ls
+      ...lists default directory.
+
+      ! ls | more
+      ...lists default directory, using more for scrolling.

analyze/ACOMAND.RUL

@@ -0,0 +1,7 @@
+* ACOMAND.RUL
+This tests ANALYZE command (from rule file).  Spec is:
+$TEXT
+ %%2 %%3 %%4 %%5 %%6 %%7 %%8 %%9
+$TEXT
+$ANALYZE  %%2 %%3 %%4 %%5 %%6 %%7 %%8 %%9
+$EXIT

analyze/ADDRIM.HLP

@@ -0,0 +1,21 @@
+  Function: Add rim elements of rows or columns
+  Syntax:   ADDRIM [{ROW | COL} [conditional]]
+
+   If ROW | COL is absent, current setting of switch DISP_COL is used.
+   If conditional is absent, current submatrix is used.
+
+    Examples:
+
+    ADDRIM R *
+    ...Adds rim of all rows
+
+    ADDR   C P STAT=B
+    ...Adds rim of all columns whose name begins with P and whose
+       solution status (S) is Basic
+    ADD   C
+    ...Adds rim of columns in submatrix
+
+    ADDR  R *NE S=LU Y=0
+    ...Adds rim of rows whose name has NE in positions 2-3, whose
+       solution status is nobasic (at Lower bound or at Upper bound),
+       and whose level (Y) is zero

analyze/AGGREGAT.DOC

@@ -0,0 +1,184 @@
+
+                              AGGREGAT Command
+
+    The AGGREGAT command computes a price-weighted sum of the rows in the
+    submatrix to produce the following single constraint (implied by
+    those in the submatrix).
+
+                        a <= Sum[j | p(j)X(j)] <= b,
+
+    where p(j) = Sum[i in submatrix | P(i)A(i,j)], and
+
+                      P(i)   = price of i-th row
+                      A(i,j) = matrix coefficient
+                      X(j)   = level of j-th activity
+
+    The aggregate coefficient, p(j), shown with the output, is formed by
+    a P-weighted sum of matrix coefficients, where P is is the dual price
+    vector.  Only the rows in the submatrix are used in forming p(j), but
+    the aggregate constraint pertains to all columns.
+
+    The aggregate limits, a and b, are:
+
+         a = Sum[i IN I+ | P(i)L(i)] + Sum[i IN I- | P(i)U(i)]
+         b = Sum[i IN I+ | P(i)U(i)] + Sum[i IN I- | P(i)L(i)],
+
+    where I+ = {i:  P(i) > 0} and I- = {i:  P(i) < 0}; and, L(i), U(i)
+    are the (original) limits on row i.  What this says is that if a
+    price is negative, the row's inequalities are reversed in forming the
+    aggregate range.
+
+    The primary purpose of the AGGREGAT command is when the LP is
+    infeasible, in which case P(i) = 0 for i = objective row; however,
+    the AGGREGAT command could be used for examining ranges of rows in an
+    optimal solution.  We shall separate these two uses of the AGGREGAT
+    command after we present a fundamental derivation.
+
+    If the submatrix contains all rows, the aggregate coefficient, p(j) =
+    P*A(j) (where A(j) includes the objective), equals minus the reduced
+    cost of activity j -- that is, P*A(j) = -D(j).  The aggregate
+    constraint, therefore, is:
+
+                      a <= y = Sum[j | -D(j)X(j)] <= b
+
+    and we require L <= X <= U.  (The variable y is introduced here for
+    notational convenience below.)
+
+                          Infeasible Linear Program
+                          ~~~~~~~~~~~~~~~~~~~~~~~~~
+    Let us begin with the infeasible case.  The output begins with a
+    statement about the "required" range.  This is [a, b] (where a = -*
+    means negative infinity and b = * means positive infinity).
+
+    Then, the nonzero aggregate coefficients are listed (see Primer for
+    discussions of WOODINFE and FOREST6, which are two infeasible linear
+    programs that came with ANALYZE).  The activities with nonzero
+    aggregate coefficients, P*A(j), are put into the submatrix; and,
+    those with zero (or nearly zero) coefficients are excluded from the
+    submatrix.  Thus, when AGGREGAT finishes, the columns in the
+    submatrix are determined by the aggregation, over-riding whatever
+    setting you had prior to the aggregation.  It is possible for all
+    coefficients to be zero, in which case there will be no columns in
+    the submatrix.
+
+    After the coefficients are listed (if any), the "myopic range" is
+    given.  This is obtained by setting each variable to one of its bound
+    values, depending upon the sign of its aggregate coefficient.  That
+    is, the range contribution from X(j) is determined as follows.
+
+    Suppose L(j) <= X(j) <= U(j), where L(j), U(j) are the (original)
+    bounds on the level, X(j) (which could be infinite).  Then, the range
+    of the term, -D(j)X(j), is given by:
+
+                 [D(j)*L(j), D(j)*U(j)]  if D(j) > 0
+                 [D(j)*U(j), D(j)*L(j)]  if D(j) < 0.
+
+    The bottom line myopic range of the aggregate row is the sum of the
+    ranges of the terms.  For an infeasible instance, with all rows in
+    the submatrix, the myopic range is necessarily disjoint from the
+    required range (displayed first).  For example, if the required range
+    is [a, b] and the myopic range is [c, d], it must be that d < a or c
+    > b, so the required range can never be satisfied.  This follows from
+    the theory of linear programming if the resident solution prices
+    represent a Phase I solution.
+
+    At a Phase I solution, the term, -D(j)X(j), has no direction of
+    change in X(j) that reduces the infeasibility.  The myopic range,
+    given at the end of the output of the AGGREGAT command, reveals the
+    range of the sum, which is disjoint from [a, b].  To aid
+    infeasibility diagnosis, the real information is in which activities
+    have nonzero aggregate coefficients.
+
+    Suppose the LP contains supplies and demands.  If d < a, this
+    suggests the aggregate demand requirement, y >= a, cannot be
+    satisfied since the myopic range means y <= d in every solution
+    satsifying L <= X <= U.  In this case, look for activities in the
+    output that limit the range from above (i.e., contributing to why d
+    is so low, relative to a).  If you find activities with positive
+    coefficients, they will have limiting upper bounds, and if you find
+    activities with negative coefficients, they will have limiting lower
+    bounds.  Sometimes, this provides enough information to form a
+    diagnosis.
+
+    Similarly, if c > b, we require y <= b, but we have y >= c.  This
+    suggests a problem with the aggregate supply limit.  The activities
+    listed (if any) might be enough information to form a diagnosis of
+    what went wrong.
+
+    To begin to understand how to use the AGGREGAT command to aid
+    infeasibility diagnosis, apply it to WOODINFE (the other instance,
+    FOREST6, is more complicated).  Details of this reasoning can be
+    found in the paper,
+
+       H.J. Greenberg, "An Empirical Analysis of Infeasible Instances
+           of a Linear Programming Blending Model," IMA Journal of
+           Mathematics Applied in Business & Industry 4 (1992), 163-210.
+
+
+                           Optimal Linear Program
+                           ~~~~~~~~~~~~~~~~~~~~~~
+    Now suppose the linear program has an optimal solution, which is
+    resident.  Then, the AGGREGAT command can be used to examine ranges
+    of rows, for the particular aggregation by optimal prices.
+
+    To help gain insight, suppose all rows are in the submatrix, so the
+    aggregate row is:
+
+                           y = Sum[j | -D(j)X(j)].
+
+    Since the objective row is included, and it is a free row (i.e., L =
+    -* and U = *), the required range is -* to *, as you will see (this
+    does not occur in an infeasible instance because then the price for
+    each free row, including the objective, is 0).
+
+    Now consider the myopic range.  For j basic, D(j) = 0, so the sum is
+    over nonbasic variables.  At optimality, if X(j) = L(j), D(j) >= 0,
+    so the range of the term, -D(j)X(j), is [-D(j)U(j), -D(j)L(j)].
+    Similarly, if X(j) = U(j), D(j) <= 0, so the range is [-D(j)L(j),
+    -D(j)U(j)].  Putting these together, the myopic range of the sum is:
+
+                  y >=  Sum[j: STAT(j) = L | -D(j)U(j)]
+                      + Sum[j: STAT(j) = U | -D(j)L(j)]
+    and
+                  y <=  Sum[j: STAT(j) = L | -D(j)L(j)]
+                      + Sum[j: STAT(j) = U | -D(j)U(j)]
+
+    If there is some activity, j, for which STAT(j) = L, D(j) > 0 and
+    U(j) = *, it follows that y >= -*.  This is typically what you will
+    see as the lower bound on the myopic range.
+
+    Now examine the upper bound on y.  It is "typical" that L >= 0 for
+    each non-basic variable, so the upper bound on y is finite.  In
+    particular, if L = 0 and U = *, the upper bound is 0.
+
+    The meaning of this case, where our LP is the canonical form,
+
+                      Minimize cx: x >= 0 and Ax >= b,
+
+    is that optimal prices weight the rows so that
+
+                  0 = Max[-DX: X >= 0] = -Min[DX: X >= 0],
+
+    which states that the objective cannot be decreased from its current
+    (optimal) level.
+
+    If you examine WOODNET, you will see another property, namely that
+    the myopic range is -* to -min.  This happens, not by coincidence,
+    but because the LP is modeled with homogeneous equations.  Its form
+    is:
+
+                      Minimize cx: L <= x <= U, Ax = 0,
+
+    where L >= 0.  Supply limits appear as upper bounds on supply
+    activities, and demand requirements appear as fixed levels, L = U > 0
+    (which could also have U = * in an equivalent formulation).  Since
+    WOODNET is a network model, the homogeneous equations, Ax = 0,
+    represent conservation of flows.
+
+    To see why y <= -min results in the myopic range, note that cx = Dx
+    since the right-hand side is 0 (from the theory of linear
+    programming, cx = Dx + Pb when LP is in standard form, where the
+    constraints (except simple bounds on x) are equations, Ax = b).
+
+    Therefore, Dx >= min, or -Dx <= -min, which is what you see in the
+    output for WOODNET.

analyze/AGGREGAT.HLP

@@ -0,0 +1,27 @@
+   Function: Display aggregation of rows in submatrix
+   Syntax:   AGGREGAT [{ROW | COL} [conditional]]
+
+   If ROW | COL is absent, current setting of switch DISP_COL is used.
+   If conditional is absent, current submatrix is used.
+
+   Aggregation weights are the resident dual variables.  This is
+   useful if the LP is infeasible and the dual prices are optimal
+   (unscaled) Phase 1 prices.  In that case, the aggregation of all
+   rows is a single infeasible constraint.
+
+   If rows are aggregated, columns with nonzero aggregate coefficients
+   are put into the submatrix;  and, columns with zero aggregate
+   coefficients are excluded from the submatrix.  (A similar result
+   occurs for column aggregation.)  See AGGREGAT.DOC (and primer) for
+   more information.
+
+   Examples:
+
+   AGGREGAT R *
+   ...Aggregates all rows
+
+   AGG R
+   ...Aggregates rows in the submatrix
+
+   AGG C * X GT L
+   ...Aggregates columns whose level is greater than its lower bound