Remove preprocessor directives from DocOnce source files

- Delete all slides directories and slide entry files (6 chapters)
- Delete index_files.do.txt
- Remove FORMAT conditionals, keeping only LaTeX/PDF content
- Remove named flags (OLD, NOTREADY, 2DO, etc.), keeping EXTRA content
- Simplify newcommands_keep.p.tex to plain LaTeX (no preprocessing)
- Update make.sh to copy instead of preprocess newcommands file
- Fix swapped label references in trunc.do.txt

Author: ggorman

doc/.src/book/book.do.txt

@@ -1,9 +1,4 @@
-# #if FORMAT == "sphinx"
-# Sphinx can only treat 63 chars in title
-TITLE:  Finite Difference Computing with PDEs
-# #else
 TITLE:  Finite Difference Computing with PDEs - A Modern Software Approach
-# #endif
 AUTHOR: Hans Petter Langtangen {copyright|CC BY} at Center for Biomedical Computing, Simula Research Laboratory & Department of Informatics, University of Oslo
 AUTHOR: Svein Linge {copyright|CC BY} at Department of Process, Energy and Environmental Technology, University College of Southeast Norway & Center for Biomedical Computing, Simula Research Laboratory
 

doc/.src/book/make.sh

@@ -44,7 +44,7 @@ if [ "$spellcheck" != 'nospell' ]; then
     fi
 fi
 
-system preprocess -DFORMAT=pdflatex ../chapters/newcommands_keep.p.tex > newcommands_keep.tex
+cp ../chapters/newcommands_keep.p.tex newcommands_keep.tex
 doconce replace 'newcommand{\E}' 'renewcommand{\E}' newcommands_keep.tex
 doconce replace 'newcommand{\I}' 'renewcommand{\I}' newcommands_keep.tex
 

doc/.src/book/preface.do.txt

@@ -197,22 +197,6 @@ the exact solution fulfilling the differential equation to the
 symbol `u` that is naturally used for the associated data structure
 in the code.
 
-# #ifdef FEM
-The mathematical notation in this book is somewhat
-different from the vast literature.
-My aim is to introduce a mathematical notation that is as
-close as possible to what we will implement in computer code.  That
-is, the way we program dictates the way we write mathematical
-formulas. The overall goal is to have as small distance as possible
-between the mathematics and the code. This is principal goal of the
-FEniCS project, and this book implements a similar goal also for
-mathematically very simple problems. Essentially, we apply a
-more classical notation when deriving methods, but then we
-get rid of unnecessary indices and other types of
-``cluttered notation'' that one would never use in code.
-The minimization of indices in the resulting formulas also help
-make the mathematical problem cleaner and clearer for students.
-# #endif
 
 === Limited scope ===
 
@@ -284,9 +268,7 @@ laboratory's financial contribution as well as to the financial
 contribution from the Department of Process, Energy and Environmental
 Technology at the University College of Southeast Norway.
 
-# #if FORMAT in ("latex", "pdflatex")
 \vspace{1cm}
 
 \noindent
 {\it Oslo, July 2016}  \hfill  {\it Hans Petter Langtangen, Svein Linge}
-# #endif

doc/.src/chapters/diffu/diffu_analysis.do.txt

@@ -930,149 +930,6 @@ Figure ref{diffu:pde1:CN:fig:F=10}. The only remedy is to lower the $F$ value.
 
 
 
-# #ifdef OLD
-#Case from Elements book where $C=0.45$ with $\Delta x=1/20, 1/60$,
-#$t=0.1$ and two clearly different plots. Wrong or correct? Formulate
-#problem. How long does it take to reach the level $u=1E-4$? Test two
-#grids, two significantly different times. Investigate with analysis.}
-#Say we simulate up to a time point $t$ and examine the amplitude:
-
-!bt
-\begin{equation*} A = \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)^{n}\end{equation*}
-!et
-
-!bt
-\begin{equation*} A = \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)^{t/\Delta t}\end{equation*}
-!et
-
-!bt
-\begin{equation*} A = \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)^{t/(C\Delta x^2)}\end{equation*}
-!et
-Expanding this in $\Delta x$ is too complicated for sympy.
-Better to plot this numerical A as function of $\Delta x$ and compare with
-the constant value from the exact expression:
-
-!bt
-\begin{equation*} A = e^{-\dfc k^2t}\end{equation*}
-!et
-FE.py plots the two and show how the numerical amplitude falls off
-as $\Delta x$ increases. Maybe this should make it into the text!
-It's yet another good example on a plot that looks strange.
-Maybe it would be good to embed the problem in a more real problem
-where something is decaying and then use the analysis to explain, using
-sine variation in space. With $k=3\pi$ we speak about a fairly smooth
-thing. There is probably a lot to explore along these lines.
-
-OK, here is the problem: some diffusion problem, big and small
-grid, simulate to a certain time, measure the level of the peak,
-see that the differences are substantial. Wrong implementation or
-a feature of the method? In the analysis we can look at FE, BE and CN.
-
-I'm not sure the rest is of value. NO!
-
-!bt
-\begin{equation*} = e^{-alpha k^2 nC\Delta x^2} = (e^{-alpha k^2 C\Delta x^2})^n\end{equation*}
-!et
-FE.py says this error is
-
-!bt
-\begin{equation*} \half C^2\Delta x^4 k^4\end{equation*}
-!et
-But it is the power $n$ that is important, it is large/small for fine/coarse
-grid.
-
-Logarithms:
-
-!bt
-\begin{equation*} \ln A = \frac{t}{C\Delta x^2}\ln \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)\end{equation*}
-!et
-
-!bt
-\begin{equation*} \ln A = -\dfc k^2t\end{equation*}
-!et
-FE.py didn't bring me to a clear answer. The error in $A$ is not enough,
-because coarse/fine grid has small/large $n$.
-
-
-
-Say we want to reduce $u$ by a factor of $r$:
-
-!bt
-\begin{equation*} \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)^n = r\end{equation*}
-!et
-
-!bt
-\begin{equation*} n = \ln r /\ln \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)\end{equation*}
-!et
-
-!bt
-\begin{equation*} t = \Delta t\ln r /\ln \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)\end{equation*}
-!et
-
-!bt
-\begin{equation*} t = C\Delta x^2\ln r /\ln \left(1 -4C\sin^2\left(\frac{k\Delta x}{2}\right)\right)\end{equation*}
-!et
-
-
-
-=== The Backward Euler Scheme ===
-
-Inserting (ref{diffu:pde1:analysis:uni})
-in
-
-!bt
-\begin{equation*} [D_t^- u = \dfc D_xD_x u]^n_j,\end{equation*}
-!et
-and dividing by $A^n e^{ikx}$ results in
-
-!bt
-\begin{equation*}
-\frac{1-A^{-1}}{\Delta t} = -\dfc \frac{4}{\Delta x^2}\sin^2\left(
-\frac{k\Delta x}{2}\right) \tp
-\end{equation*}
-!et
-Solving with respect to $A$ gives an amplification factor of
-
-
-=== The $\theta$ Rule ===
-
-The analysis of the
-$\theta$ rule becomes easy if we express the scheme
-as
-
-!bt
-\begin{equation*} [D_t u]^{n+\halfi}_j = \dfc\theta [D_xD_x]^{n}_j +
-\dfc(1-\theta)[D_xD_x]^{n-1}_j \tp  \end{equation*}
-!et
-Inserting (ref{diffu:pde1:analysis:uni}) and dividing by
-$A^n e^{ikx}$ gives the equation
-
-!bt
-\begin{equation*} \frac{A-1}{\Delta t} =
-\dfc 4C\sin^2\left(
-\frac{k\Delta x}{2}\right)\left( \theta A +
-(1-\theta)\right),\end{equation*}
-!et
-which leads to
-
-!bt
-\begin{equation}
-A = \frac{1 - \dfc (1-\theta)}{1+\dfc\theta}
-C\sin^2\left(
-\frac{k\Delta x}{2}\right),
-\end{equation}
-!et
-and
-
-!bt
-\begin{equation}
-u^n_j = \left(\frac{1 - \dfc (1-\theta)}{1+\dfc\theta}
-C\sin^2\left(
-\frac{k\Delta x}{2}\right)\right)^n e^{ikj\Delta x} \tp  \end{equation}
-!et
-
-Use these in tests....
-# #endif
 
 ======= Exercises =======
 

doc/.src/chapters/diffu/diffu_fd1.do.txt

@@ -94,7 +94,6 @@ We shall start with the simplest boundary condition: $u=0$. The
 complete initial-boundary value diffusion problem in one space
 dimension can then be specified as
 
-# #if FORMAT in ("latex", "pdflatex")
 !bt
 \begin{alignat}{3}
 \frac{\partial u}{\partial t} &=
@@ -109,21 +108,6 @@ u(L,t) & = 0, \quad & t>0
 label{diffu:pde1:bc:L}
 \end{alignat}
 !et
-# #else
-!bt
-\begin{align}
-\frac{\partial u}{\partial t} &=
-\dfc \frac{\partial^2 u}{\partial x^2} + f, \quad x\in (0,L),\ t\in (0,T]
-label{diffu:pde1}\\
-u(x,0) &= I(x), \quad  x\in [0,L]
-label{diffu:pde1:ic:u}\\
-u(0,t) & = 0, \quad  t>0,
-label{diffu:pde1:bc:0}\\
-u(L,t) & = 0, \quad  t>0\tp
-label{diffu:pde1:bc:L}
-\end{align}
-!et
-# #endif
 With only a first-order derivative in time,
 only one *initial condition* is needed, while the second-order
 derivative in space leads to a demand for two *boundary conditions*.
@@ -1137,13 +1121,11 @@ be first order convergence, whereas with $\theta=0.5$ (i.e., Crank-Nicolson), on
 second order convergence (as outlined in previous sections).
 
 
-# #ifdef EXTRA
 !bt
 \begin{equation}
 [D_t u = \dfc D_xD_x \overline{u}^{t,\theta}]^n_i
 \end{equation}
 !et
-# #endif
 
 ===== Experiments =====
 label{diffu:pde1:theta:experiments}
@@ -1217,121 +1199,3 @@ pseudo time variable.
 
 [hpl: Better to say the last sentence when we treat iterative methods.]
 
-# #ifdef 2DO
-
-===== Extensions =====
-
-These extensions are performed exactly as for a wave equation as they
-only affect the spatial derivatives (which are the same as in the
-wave equation).
-
- * Variable coefficients
- * Neumann and Robin conditions
- * 2D and 3D
-
-Future versions of this document will for completeness and
-independence of the wave equation document feature info on the three
-points. The Robin condition is new, but straightforward to handle:
-
-!bt
-\[ -\dfc\frac{\partial u}{\partial n} = h_T(u-U_s),\quad
-[-\dfc D_x u = h_T(u-U_s)]^n_i
-\]
-!et
-
-===== Implementation =====
-label{diffu:pde1:impl}
-
-
-===== Adaptive Time Stepping =====
-
-Intuitive approach based on strategies from ODE decay equation and
-a very discrete Fourier composition of analytical components.
-What about $||u_t|| < \epsilon_L \Rightarrow \Delta t$ doubled,
-$||u_t|| > \epsilon_U \Rightarrow \Delta t$ halved? Google...
-
-
-
-===== Variable Diffusion Coefficient =====
-
-
-!bt
-\begin{equation}
-\frac{\partial u}{\partial t} =
-\frac{\partial}{\partial x}\left(
-\dfc(x) \frac{\partial u}{\partial x}\right)
-\end{equation}
-!et
-
-It is always very useful to have analytical solutions to PDEs.
-For the variable coefficient diffusion equation this is difficult
-to find unless we add a source term and apply the method of
-manufactured solutions. In the  limit $t\rightarrow\infty$, however,
-we can derive an analytical solution.
-The limit problem for the stationary solution
-$v(x) = \lim_{t\rightarrow \infty} u(x,t)$,
-assuming $\partial u/\partial t\rightarrow 0$ in this limit, becomes
-
-!bt
-\begin{equation}
-\frac{\partial}{\partial x}\left(
-\dfc(x) \frac{\partial u}{\partial x}\right)
-= 0,\quad v(0)=u_L,\ v(L)=u_R \tp  \end{equation}
-!et
-Integrating twice results in
-
-!bt
-\begin{equation*} v = C_2 + C_1\int_0^x \frac{1}{\dfc(\xi)} d\xi \tp  \end{equation*}
-!et
-The two integration constants $C_1$ and $C_2$ are determined from
-the boundary conditions, and $v(x)$ becomes
-
-!bt
-\begin{equation}
-v(x) = u_L + (u_R - u_L)
-\frac{\int_0^x\frac{d\xi}{a(\xi)}}{\int_0^L\frac{d\xi}{a(\xi)}}
-\end{equation}
-!et
-
-Evaluation of $v$ on a mesh with points $x_i$, $i=0,\ldots,N_x$,
-can be performed using the Trapezoidal rule.
-Defining $g(x) = \int_0^x (\dfc(\xi))^{-1}d\xi$, $g_j$ can be
-computed as
-
-!bt
-\begin{equation}
-g_i = \left\lbrace\begin{array}{ll}
-\half \Delta x /\dfc_i, & i = 0,\\
-g_{i-1} + \Delta x /\dfc_i, & 0 < i < N_x,\\
-g_{i-1} + \half\Delta x /\dfc_i, & i = N_x \tp  \end{array}\right.
-\end{equation}
-!et
-Then
-
-!bt
-\begin{equation}
-v_i = u_L + (u_R - u_L)g_i/g_{N_x} \tp  \end{equation}
-!et
-The corresponding implementation can be written
-
-!bc pycod
-def u_exact_stationary(x, a, u_L, u_R):
-    Nx = x.size - 1
-    g = zeros(Nx+1)    # integral of 1/a from 0 to x
-    dx = x[1] - x[0]   # assumed constant
-    i = 0
-    g[i] = 0.5*dx/a[i]
-    for i in range(1, Nx):
-        g[i] = g[i-1] + dx/a[i]
-    i = Nx
-    g[i] = g[i-1] + 0.5*dx/a[i]
-    v = u_L + (u_R - u_L)*g/g[-1]
-    return v
-!ec
-
-
-
-======= Two-Dimensional Diffusion =======
-label{diffu:2D}
-
-# #endif