Import Ghostscript 9.54ghostscript-9.54

Signed-off-by: Thomas Deutschmann <whissi@gentoo.org>

1 files changed, 1097 insertions, 0 deletions

diff --git a/extract/src/docx.c b/extract/src/docx.c
new file mode 100644
index 00000000..238e81d4
--- /dev/null
+++ b/extract/src/docx.c
@@ -0,0 +1,1097 @@
+/* These extract_docx_*() functions generate docx content and docx zip archive
+data.
+
+Caller must call things in a sensible order to create valid content -
+e.g. don't call docx_paragraph_start() twice without intervening call to
+docx_paragraph_finish(). */
+
+#include "../include/extract.h"
+
+#include "docx_template.h"
+
+#include "astring.h"
+#include "document.h"
+#include "docx.h"
+#include "mem.h"
+#include "memento.h"
+#include "outf.h"
+#include "zip.h"
+
+#include <assert.h>
+#include <errno.h>
+#include <math.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+
+#include <sys/stat.h>
+
+
+static int extract_docx_paragraph_start(extract_alloc_t* alloc, extract_astring_t* content)
+{
+    return extract_astring_cat(alloc, content, "\n\n<w:p>");
+}
+
+static int extract_docx_paragraph_finish(extract_alloc_t* alloc, extract_astring_t* content)
+{
+    return extract_astring_cat(alloc, content, "\n</w:p>");
+}
+
+static int extract_docx_run_start(
+        extract_alloc_t* alloc,
+        extract_astring_t* content,
+        const char* font_name,
+        double font_size,
+        int bold,
+        int italic
+        )
+/* Starts a new run. Caller must ensure that extract_docx_run_finish() was
+called to terminate any previous run. */
+{
+    int e = 0;
+    if (!e) e = extract_astring_cat(alloc, content, "\n<w:r><w:rPr><w:rFonts w:ascii=\"");
+    if (!e) e = extract_astring_cat(alloc, content, font_name);
+    if (!e) e = extract_astring_cat(alloc, content, "\" w:hAnsi=\"");
+    if (!e) e = extract_astring_cat(alloc, content, font_name);
+    if (!e) e = extract_astring_cat(alloc, content, "\"/>");
+    if (!e && bold) e = extract_astring_cat(alloc, content, "<w:b/>");
+    if (!e && italic) e = extract_astring_cat(alloc, content, "<w:i/>");
+    {
+        char   font_size_text[32];
+        if (0) font_size = 10;
+
+        if (!e) e = extract_astring_cat(alloc, content, "<w:sz w:val=\"");
+        snprintf(font_size_text, sizeof(font_size_text), "%f", font_size * 2);
+        extract_astring_cat(alloc, content, font_size_text);
+        extract_astring_cat(alloc, content, "\"/>");
+
+        if (!e) e = extract_astring_cat(alloc, content, "<w:szCs w:val=\"");
+        snprintf(font_size_text, sizeof(font_size_text), "%f", font_size * 1.5);
+        extract_astring_cat(alloc, content, font_size_text);
+        extract_astring_cat(alloc, content, "\"/>");
+    }
+    if (!e) e = extract_astring_cat(alloc, content, "</w:rPr><w:t xml:space=\"preserve\">");
+    return e;
+
+}
+
+static int extract_docx_run_finish(extract_alloc_t* alloc, extract_astring_t* content)
+{
+    return extract_astring_cat(alloc, content, "</w:t></w:r>");
+}
+
+static int extract_docx_char_append_string(extract_alloc_t* alloc, extract_astring_t* content, const char* text)
+{
+    return extract_astring_cat(alloc, content, text);
+}
+
+static int extract_docx_char_append_stringf(extract_alloc_t* alloc, extract_astring_t* content, const char* format, ...)
+{
+    char* buffer = NULL;
+    int e;
+    va_list va;
+    va_start(va, format);
+    e = extract_vasprintf(alloc, &buffer, format, va);
+    va_end(va);
+    if (e < 0) return e;
+    e = extract_astring_cat(alloc, content, buffer);
+    extract_free(alloc, &buffer);
+    return e;
+}
+
+static int extract_docx_char_append_char(extract_alloc_t* alloc, extract_astring_t* content, char c)
+{
+    return extract_astring_catc(alloc, content, c);
+}
+
+static int extract_docx_paragraph_empty(extract_alloc_t* alloc, extract_astring_t* content)
+/* Append an empty paragraph to *content. */
+{
+    int e = -1;
+    if (extract_docx_paragraph_start(alloc, content)) goto end;
+    /* It seems like our choice of font size here doesn't make any difference
+    to the ammount of vertical space, unless we include a non-space
+    character. Presumably something to do with the styles in the template
+    document. */
+    if (extract_docx_run_start(
+            alloc,
+            content,
+            "OpenSans",
+            10 /*font_size*/,
+            0 /*font_bold*/,
+            0 /*font_italic*/
+            )) goto end;
+    //docx_char_append_string(content, "&#160;");   /* &#160; is non-break space. */
+    if (extract_docx_run_finish(alloc, content)) goto end;
+    if (extract_docx_paragraph_finish(alloc, content)) goto end;
+    e = 0;
+    end:
+    return e;
+}
+
+
+/* Removes last <len> chars. */
+static int docx_char_truncate(extract_astring_t* content, int len)
+{
+    assert((size_t) len <= content->chars_num);
+    content->chars_num -= len;
+    content->chars[content->chars_num] = 0;
+    return 0;
+}
+
+static int extract_docx_char_truncate_if(extract_astring_t* content, char c)
+/* Removes last char if it is <c>. */
+{
+    if (content->chars_num && content->chars[content->chars_num-1] == c) {
+        docx_char_truncate(content, 1);
+    }
+    return 0;
+}
+
+
+static double matrices_to_font_size(matrix_t* ctm, matrix_t* trm)
+{
+    double font_size = matrix_expansion(*trm)
+            * matrix_expansion(*ctm);
+    /* Round font_size to nearest 0.01. */
+    font_size = (double) (int) (font_size * 100.0f + 0.5f) / 100.0f;
+    return font_size;
+}
+
+typedef struct
+{
+    const char* font_name;
+    double      font_size;
+    int         font_bold;
+    int         font_italic;
+    matrix_t*   ctm_prev;
+} content_state_t;
+/* Used to keep track of font information when writing paragraphs of docx
+content, e.g. so we know whether a font has changed so need to start a new docx
+span. */
+
+
+static int extract_document_to_docx_content_paragraph(
+        extract_alloc_t*    alloc,
+        content_state_t*    state,
+        paragraph_t*        paragraph,
+        extract_astring_t*  content
+        )
+/* Append docx xml for <paragraph> to <content>. Updates *state if we change
+font. */
+{
+    int e = -1;
+    int l;
+    if (extract_docx_paragraph_start(alloc, content)) goto end;
+
+    for (l=0; l<paragraph->lines_num; ++l) {
+        line_t* line = paragraph->lines[l];
+        int s;
+        for (s=0; s<line->spans_num; ++s) {
+            int si;
+            span_t* span = line->spans[s];
+            double font_size_new;
+            state->ctm_prev = &span->ctm;
+            font_size_new = matrices_to_font_size(&span->ctm, &span->trm);
+            if (!state->font_name
+                    || strcmp(span->font_name, state->font_name)
+                    || span->font_bold != state->font_bold
+                    || span->font_italic != state->font_italic
+                    || font_size_new != state->font_size
+                    ) {
+                if (state->font_name) {
+                    if (extract_docx_run_finish(alloc, content)) goto end;
+                }
+                state->font_name = span->font_name;
+                state->font_bold = span->font_bold;
+                state->font_italic = span->font_italic;
+                state->font_size = font_size_new;
+                if (extract_docx_run_start(
+                        alloc,
+                        content,
+                        state->font_name,
+                        state->font_size,
+                        state->font_bold,
+                        state->font_italic
+                        )) goto end;
+            }
+
+            for (si=0; si<span->chars_num; ++si) {
+                char_t* char_ = &span->chars[si];
+                int c = char_->ucs;
+
+                if (0) {}
+
+                /* Escape XML special characters. */
+                else if (c == '<')  extract_docx_char_append_string(alloc, content, "&lt;");
+                else if (c == '>')  extract_docx_char_append_string(alloc, content, "&gt;");
+                else if (c == '&')  extract_docx_char_append_string(alloc, content, "&amp;");
+                else if (c == '"')  extract_docx_char_append_string(alloc, content, "&quot;");
+                else if (c == '\'') extract_docx_char_append_string(alloc, content, "&apos;");
+
+                /* Expand ligatures. */
+                else if (c == 0xFB00) {
+                    if (extract_docx_char_append_string(alloc, content, "ff")) goto end;
+                }
+                else if (c == 0xFB01) {
+                    if (extract_docx_char_append_string(alloc, content, "fi")) goto end;
+                }
+                else if (c == 0xFB02) {
+                    if (extract_docx_char_append_string(alloc, content, "fl")) goto end;
+                }
+                else if (c == 0xFB03) {
+                    if (extract_docx_char_append_string(alloc, content, "ffi")) goto end;
+                }
+                else if (c == 0xFB04) {
+                    if (extract_docx_char_append_string(alloc, content, "ffl")) goto end;
+                }
+
+                /* Output ASCII verbatim. */
+                else if (c >= 32 && c <= 127) {
+                    if (extract_docx_char_append_char(alloc, content, (char) c)) goto end;
+                }
+
+                /* Escape all other characters. */
+                else {
+                    char    buffer[32];
+                    snprintf(buffer, sizeof(buffer), "&#x%x;", c);
+                    if (extract_docx_char_append_string(alloc, content, buffer)) goto end;
+                }
+            }
+            /* Remove any trailing '-' at end of line. */
+            if (extract_docx_char_truncate_if(content, '-')) goto end;
+        }
+    }
+    if (state->font_name) {
+        if (extract_docx_run_finish(alloc, content)) goto end;
+        state->font_name = NULL;
+    }
+    if (extract_docx_paragraph_finish(alloc, content)) goto end;
+    
+    e = 0;
+    
+    end:
+    return e;
+}
+
+static int extract_document_append_image(
+        extract_alloc_t*    alloc,
+        extract_astring_t*  content,
+        image_t*            image
+        )
+/* Write reference to image into docx content. */
+{
+    extract_docx_char_append_string(alloc, content, "\n");
+    extract_docx_char_append_string(alloc, content, "     <w:p>\n");
+    extract_docx_char_append_string(alloc, content, "       <w:r>\n");
+    extract_docx_char_append_string(alloc, content, "         <w:rPr>\n");
+    extract_docx_char_append_string(alloc, content, "           <w:noProof/>\n");
+    extract_docx_char_append_string(alloc, content, "         </w:rPr>\n");
+    extract_docx_char_append_string(alloc, content, "         <w:drawing>\n");
+    extract_docx_char_append_string(alloc, content, "           <wp:inline distT=\"0\" distB=\"0\" distL=\"0\" distR=\"0\" wp14:anchorId=\"7057A832\" wp14:editId=\"466EB3FB\">\n");
+    extract_docx_char_append_string(alloc, content, "             <wp:extent cx=\"2933700\" cy=\"2200275\"/>\n");
+    extract_docx_char_append_string(alloc, content, "             <wp:effectExtent l=\"0\" t=\"0\" r=\"0\" b=\"9525\"/>\n");
+    extract_docx_char_append_string(alloc, content, "             <wp:docPr id=\"1\" name=\"Picture 1\"/>\n");
+    extract_docx_char_append_string(alloc, content, "             <wp:cNvGraphicFramePr>\n");
+    extract_docx_char_append_string(alloc, content, "               <a:graphicFrameLocks xmlns:a=\"http://schemas.openxmlformats.org/drawingml/2006/main\" noChangeAspect=\"1\"/>\n");
+    extract_docx_char_append_string(alloc, content, "             </wp:cNvGraphicFramePr>\n");
+    extract_docx_char_append_string(alloc, content, "             <a:graphic xmlns:a=\"http://schemas.openxmlformats.org/drawingml/2006/main\">\n");
+    extract_docx_char_append_string(alloc, content, "               <a:graphicData uri=\"http://schemas.openxmlformats.org/drawingml/2006/picture\">\n");
+    extract_docx_char_append_string(alloc, content, "                 <pic:pic xmlns:pic=\"http://schemas.openxmlformats.org/drawingml/2006/picture\">\n");
+    extract_docx_char_append_string(alloc, content, "                   <pic:nvPicPr>\n");
+    extract_docx_char_append_string(alloc, content, "                     <pic:cNvPr id=\"1\" name=\"Picture 1\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                     <pic:cNvPicPr>\n");
+    extract_docx_char_append_string(alloc, content, "                       <a:picLocks noChangeAspect=\"1\" noChangeArrowheads=\"1\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                     </pic:cNvPicPr>\n");
+    extract_docx_char_append_string(alloc, content, "                   </pic:nvPicPr>\n");
+    extract_docx_char_append_string(alloc, content, "                   <pic:blipFill>\n");
+    extract_docx_char_append_stringf(alloc, content,"                     <a:blip r:embed=\"%s\">\n", image->id);
+    extract_docx_char_append_string(alloc, content, "                       <a:extLst>\n");
+    extract_docx_char_append_string(alloc, content, "                         <a:ext uri=\"{28A0092B-C50C-407E-A947-70E740481C1C}\">\n");
+    extract_docx_char_append_string(alloc, content, "                           <a14:useLocalDpi xmlns:a14=\"http://schemas.microsoft.com/office/drawing/2010/main\" val=\"0\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                         </a:ext>\n");
+    extract_docx_char_append_string(alloc, content, "                       </a:extLst>\n");
+    extract_docx_char_append_string(alloc, content, "                     </a:blip>\n");
+    //extract_docx_char_append_string(alloc, content, "                     <a:srcRect/>\n");
+    extract_docx_char_append_string(alloc, content, "                     <a:stretch>\n");
+    extract_docx_char_append_string(alloc, content, "                       <a:fillRect/>\n");
+    extract_docx_char_append_string(alloc, content, "                     </a:stretch>\n");
+    extract_docx_char_append_string(alloc, content, "                   </pic:blipFill>\n");
+    extract_docx_char_append_string(alloc, content, "                   <pic:spPr bwMode=\"auto\">\n");
+    extract_docx_char_append_string(alloc, content, "                     <a:xfrm>\n");
+    extract_docx_char_append_string(alloc, content, "                       <a:off x=\"0\" y=\"0\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                       <a:ext cx=\"2933700\" cy=\"2200275\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                     </a:xfrm>\n");
+    extract_docx_char_append_string(alloc, content, "                     <a:prstGeom prst=\"rect\">\n");
+    extract_docx_char_append_string(alloc, content, "                       <a:avLst/>\n");
+    extract_docx_char_append_string(alloc, content, "                     </a:prstGeom>\n");
+    extract_docx_char_append_string(alloc, content, "                     <a:noFill/>\n");
+    extract_docx_char_append_string(alloc, content, "                     <a:ln>\n");
+    extract_docx_char_append_string(alloc, content, "                       <a:noFill/>\n");
+    extract_docx_char_append_string(alloc, content, "                     </a:ln>\n");
+    extract_docx_char_append_string(alloc, content, "                   </pic:spPr>\n");
+    extract_docx_char_append_string(alloc, content, "                 </pic:pic>\n");
+    extract_docx_char_append_string(alloc, content, "               </a:graphicData>\n");
+    extract_docx_char_append_string(alloc, content, "             </a:graphic>\n");
+    extract_docx_char_append_string(alloc, content, "           </wp:inline>\n");
+    extract_docx_char_append_string(alloc, content, "         </w:drawing>\n");
+    extract_docx_char_append_string(alloc, content, "       </w:r>\n");
+    extract_docx_char_append_string(alloc, content, "     </w:p>\n");
+    extract_docx_char_append_string(alloc, content, "\n");
+    return 0;
+}
+
+
+static int extract_document_output_rotated_paragraphs(
+        extract_alloc_t*    alloc,
+        page_t*             page,
+        int                 paragraph_begin,
+        int                 paragraph_end,
+        int                 rot,
+        int                 x,
+        int                 y,
+        int                 w,
+        int                 h,
+        int                 text_box_id,
+        extract_astring_t*  content,
+        content_state_t*    state
+        )
+/* Writes paragraph to content inside rotated text box. */
+{
+    int e = 0;
+    int p;
+    outf("x,y=%ik,%ik = %i,%i", x/1000, y/1000, x, y);
+    extract_docx_char_append_string(alloc, content, "\n");
+    extract_docx_char_append_string(alloc, content, "\n");
+    extract_docx_char_append_string(alloc, content, "<w:p>\n");
+    extract_docx_char_append_string(alloc, content, "  <w:r>\n");
+    extract_docx_char_append_string(alloc, content, "    <mc:AlternateContent>\n");
+    extract_docx_char_append_string(alloc, content, "      <mc:Choice Requires=\"wps\">\n");
+    extract_docx_char_append_string(alloc, content, "        <w:drawing>\n");
+    extract_docx_char_append_string(alloc, content, "          <wp:anchor distT=\"0\" distB=\"0\" distL=\"0\" distR=\"0\" simplePos=\"0\" relativeHeight=\"0\" behindDoc=\"0\" locked=\"0\" layoutInCell=\"1\" allowOverlap=\"1\" wp14:anchorId=\"53A210D1\" wp14:editId=\"2B7E8016\">\n");
+    extract_docx_char_append_string(alloc, content, "            <wp:simplePos x=\"0\" y=\"0\"/>\n");
+    extract_docx_char_append_string(alloc, content, "            <wp:positionH relativeFrom=\"page\">\n");
+    extract_docx_char_append_stringf(alloc, content,"              <wp:posOffset>%i</wp:posOffset>\n", x);
+    extract_docx_char_append_string(alloc, content, "            </wp:positionH>\n");
+    extract_docx_char_append_string(alloc, content, "            <wp:positionV relativeFrom=\"page\">\n");
+    extract_docx_char_append_stringf(alloc, content,"              <wp:posOffset>%i</wp:posOffset>\n", y);
+    extract_docx_char_append_string(alloc, content, "            </wp:positionV>\n");
+    extract_docx_char_append_stringf(alloc, content,"            <wp:extent cx=\"%i\" cy=\"%i\"/>\n", w, h);
+    extract_docx_char_append_string(alloc, content, "            <wp:effectExtent l=\"381000\" t=\"723900\" r=\"371475\" b=\"723900\"/>\n");
+    extract_docx_char_append_string(alloc, content, "            <wp:wrapNone/>\n");
+    extract_docx_char_append_stringf(alloc, content,"            <wp:docPr id=\"%i\" name=\"Text Box %i\"/>\n", text_box_id, text_box_id);
+    extract_docx_char_append_string(alloc, content, "            <wp:cNvGraphicFramePr/>\n");
+    extract_docx_char_append_string(alloc, content, "            <a:graphic xmlns:a=\"http://schemas.openxmlformats.org/drawingml/2006/main\">\n");
+    extract_docx_char_append_string(alloc, content, "              <a:graphicData uri=\"http://schemas.microsoft.com/office/word/2010/wordprocessingShape\">\n");
+    extract_docx_char_append_string(alloc, content, "                <wps:wsp>\n");
+    extract_docx_char_append_string(alloc, content, "                  <wps:cNvSpPr txBox=\"1\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                  <wps:spPr>\n");
+    extract_docx_char_append_stringf(alloc, content,"                    <a:xfrm rot=\"%i\">\n", rot);
+    extract_docx_char_append_string(alloc, content, "                      <a:off x=\"0\" y=\"0\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                      <a:ext cx=\"3228975\" cy=\"2286000\"/>\n");
+    extract_docx_char_append_string(alloc, content, "                    </a:xfrm>\n");
+    extract_docx_char_append_string(alloc, content, "                    <a:prstGeom prst=\"rect\">\n");
+    extract_docx_char_append_string(alloc, content, "                      <a:avLst/>\n");
+    extract_docx_char_append_string(alloc, content, "                    </a:prstGeom>\n");
+
+    /* Give box a solid background. */
+    if (0) {
+        extract_docx_char_append_string(alloc, content, "                    <a:solidFill>\n");
+        extract_docx_char_append_string(alloc, content, "                      <a:schemeClr val=\"lt1\"/>\n");
+        extract_docx_char_append_string(alloc, content, "                    </a:solidFill>\n");
+        }
+
+    /* Draw line around box. */
+    if (0) {
+        extract_docx_char_append_string(alloc, content, "                    <a:ln w=\"175\">\n");
+        extract_docx_char_append_string(alloc, content, "                      <a:solidFill>\n");
+        extract_docx_char_append_string(alloc, content, "                        <a:prstClr val=\"black\"/>\n");
+        extract_docx_char_append_string(alloc, content, "                      </a:solidFill>\n");
+        extract_docx_char_append_string(alloc, content, "                    </a:ln>\n");
+    }
+
+    extract_docx_char_append_string(alloc, content, "                  </wps:spPr>\n");
+    extract_docx_char_append_string(alloc, content, "                  <wps:txbx>\n");
+    extract_docx_char_append_string(alloc, content, "                    <w:txbxContent>");
+
+    #if 0
+    if (0) {
+        /* Output inline text describing the rotation. */
+        extract_docx_char_append_stringf(content, "<w:p>\n"
+                "<w:r><w:rPr><w:rFonts w:ascii=\"OpenSans\" w:hAnsi=\"OpenSans\"/><w:sz w:val=\"20.000000\"/><w:szCs w:val=\"15.000000\"/></w:rPr><w:t xml:space=\"preserve\">*** rotate: %f rad, %f deg. rot=%i</w:t></w:r>\n"
+                "</w:p>\n",
+                rotate,
+                rotate * 180 / pi,
+                rot
+                );
+    }
+    #endif
+
+    /* Output paragraphs p0..p2-1. */
+    for (p=paragraph_begin; p<paragraph_end; ++p) {
+        paragraph_t* paragraph = page->paragraphs[p];
+        if (extract_document_to_docx_content_paragraph(alloc, state, paragraph, content)) goto end;
+    }
+
+    extract_docx_char_append_string(alloc, content, "\n");
+    extract_docx_char_append_string(alloc, content, "                    </w:txbxContent>\n");
+    extract_docx_char_append_string(alloc, content, "                  </wps:txbx>\n");
+    extract_docx_char_append_string(alloc, content, "                  <wps:bodyPr rot=\"0\" spcFirstLastPara=\"0\" vertOverflow=\"overflow\" horzOverflow=\"overflow\" vert=\"horz\" wrap=\"square\" lIns=\"91440\" tIns=\"45720\" rIns=\"91440\" bIns=\"45720\" numCol=\"1\" spcCol=\"0\" rtlCol=\"0\" fromWordArt=\"0\" anchor=\"t\" anchorCtr=\"0\" forceAA=\"0\" compatLnSpc=\"1\">\n");
+    extract_docx_char_append_string(alloc, content, "                    <a:prstTxWarp prst=\"textNoShape\">\n");
+    extract_docx_char_append_string(alloc, content, "                      <a:avLst/>\n");
+    extract_docx_char_append_string(alloc, content, "                    </a:prstTxWarp>\n");
+    extract_docx_char_append_string(alloc, content, "                    <a:noAutofit/>\n");
+    extract_docx_char_append_string(alloc, content, "                  </wps:bodyPr>\n");
+    extract_docx_char_append_string(alloc, content, "                </wps:wsp>\n");
+    extract_docx_char_append_string(alloc, content, "              </a:graphicData>\n");
+    extract_docx_char_append_string(alloc, content, "            </a:graphic>\n");
+    extract_docx_char_append_string(alloc, content, "          </wp:anchor>\n");
+    extract_docx_char_append_string(alloc, content, "        </w:drawing>\n");
+    extract_docx_char_append_string(alloc, content, "      </mc:Choice>\n");
+
+    /* This fallback is copied from a real Word document. Not sure
+    whether it works - both Libreoffice and Word use the above
+    choice. */
+    extract_docx_char_append_string(alloc, content, "      <mc:Fallback>\n");
+    extract_docx_char_append_string(alloc, content, "        <w:pict>\n");
+    extract_docx_char_append_string(alloc, content, "          <v:shapetype w14:anchorId=\"53A210D1\" id=\"_x0000_t202\" coordsize=\"21600,21600\" o:spt=\"202\" path=\"m,l,21600r21600,l21600,xe\">\n");
+    extract_docx_char_append_string(alloc, content, "            <v:stroke joinstyle=\"miter\"/>\n");
+    extract_docx_char_append_string(alloc, content, "            <v:path gradientshapeok=\"t\" o:connecttype=\"rect\"/>\n");
+    extract_docx_char_append_string(alloc, content, "          </v:shapetype>\n");
+    extract_docx_char_append_stringf(alloc, content,"          <v:shape id=\"Text Box %i\" o:spid=\"_x0000_s1026\" type=\"#_x0000_t202\" style=\"position:absolute;margin-left:71.25pt;margin-top:48.75pt;width:254.25pt;height:180pt;rotation:-2241476fd;z-index:251659264;visibility:visible;mso-wrap-style:square;mso-wrap-distance-left:9pt;mso-wrap-distance-top:0;mso-wrap-distance-right:9pt;mso-wrap-distance-bottom:0;mso-position-horizontal:absolute;mso-position-horizontal-relative:text;mso-position-vertical:absolute;mso-position-vertical-relative:text;v-text-anchor:top\" o:gfxdata=\"UEsDBBQABgAIAAAAIQC2gziS/gAAAOEBAAATAAAAW0NvbnRlbnRfVHlwZXNdLnhtbJSRQU7DMBBF&#10;90jcwfIWJU67QAgl6YK0S0CoHGBkTxKLZGx5TGhvj5O2G0SRWNoz/78nu9wcxkFMGNg6quQqL6RA&#10;0s5Y6ir5vt9lD1JwBDIwOMJKHpHlpr69KfdHjyxSmriSfYz+USnWPY7AufNIadK6MEJMx9ApD/oD&#10;OlTrorhX2lFEilmcO2RdNtjC5xDF9pCuTyYBB5bi6bQ4syoJ3g9WQ0ymaiLzg5KdCXlKLjvcW893&#10;SUOqXwnz5DrgnHtJTxOsQfEKIT7DmDSUCaxw7Rqn8787ZsmRM9e2VmPeBN4uqYvTtW7jvijg9N/y&#10;JsXecLq0q+WD6m8AAAD//wMAUEsDBBQABgAIAAAAIQA4/SH/1gAAAJQBAAALAAAAX3JlbHMvLnJl&#10;bHOkkMFqwzAMhu+DvYPRfXGawxijTi+j0GvpHsDYimMaW0Yy2fr2M4PBMnrbUb/Q94l/f/hMi1qR&#10;JVI2sOt6UJgd+ZiDgffL8ekFlFSbvV0oo4EbChzGx4f9GRdb25HMsYhqlCwG5lrLq9biZkxWOiqY&#10;22YiTra2kYMu1l1tQD30/bPm3wwYN0x18gb45AdQl1tp5j/sFB2T0FQ7R0nTNEV3j6o9feQzro1i&#10;OWA14Fm+Q8a1a8+Bvu/d/dMb2JY5uiPbhG/ktn4cqGU/er3pcvwCAAD//wMAUEsDBBQABgAIAAAA&#10;IQDQg5pQVgIAALEEAAAOAAAAZHJzL2Uyb0RvYy54bWysVE1v2zAMvQ/YfxB0X+2k+WiDOEXWosOA&#10;oi3QDj0rstwYk0VNUmJ3v35PipMl3U7DLgJFPj+Rj6TnV12j2VY5X5Mp+OAs50wZSWVtXgv+7fn2&#10;0wVnPghTCk1GFfxNeX61+Phh3tqZGtKadKkcA4nxs9YWfB2CnWWZl2vVCH9GVhkEK3KNCLi616x0&#10;ogV7o7Nhnk+yllxpHUnlPbw3uyBfJP6qUjI8VJVXgemCI7eQTpfOVTyzxVzMXp2w61r2aYh/yKIR&#10;tcGjB6obEQTbuPoPqqaWjjxV4UxSk1FV1VKlGlDNIH9XzdNaWJVqgTjeHmTy/49W3m8fHatL9I4z&#10;Ixq06Fl1gX2mjg2iOq31M4CeLGChgzsie7+HMxbdVa5hjiDu4HI8ml5MpkkLVMcAh+xvB6kjt4Tz&#10;fDi8uJyOOZOIwZ7keWpGtmOLrNb58EVRw6JRcIdeJlqxvfMBGQC6h0S4J12Xt7XW6RLnR11rx7YC&#10;ndch5YwvTlDasLbgk/NxnohPYpH68P1KC/k9Vn3KgJs2cEaNdlpEK3SrrhdoReUbdEvSQAZv5W0N&#10;3jvhw6NwGDQ4sTzhAUelCclQb3G2Jvfzb/6IR/8R5azF4Bbc/9gIpzjTXw0m43IwGsVJT5fReDrE&#10;xR1HVscRs2muCQqh+8gumREf9N6sHDUv2LFlfBUhYSTeLnjYm9dht07YUamWywTCbFsR7syTlZF6&#10;383n7kU42/czYBTuaT/iYvaurTts/NLQchOoqlPPo8A7VXvdsRepLf0Ox8U7vifU7z/N4hcAAAD/&#10;/wMAUEsDBBQABgAIAAAAIQBh17L63wAAAAoBAAAPAAAAZHJzL2Rvd25yZXYueG1sTI9BT4NAEIXv&#10;Jv6HzZh4s0ubgpayNIboSW3Syg9Y2BGI7CyyS0v99Y4nPU3ezMub72W72fbihKPvHClYLiIQSLUz&#10;HTUKyvfnuwcQPmgyuneECi7oYZdfX2U6Ne5MBzwdQyM4hHyqFbQhDKmUvm7Rar9wAxLfPtxodWA5&#10;NtKM+szhtperKEqk1R3xh1YPWLRYfx4nq8APVfz9VQxPb+WUNC+vZbGPDhelbm/mxy2IgHP4M8Mv&#10;PqNDzkyVm8h40bNer2K2Ktjc82RDEi+5XKVgHfNG5pn8XyH/AQAA//8DAFBLAQItABQABgAIAAAA&#10;IQC2gziS/gAAAOEBAAATAAAAAAAAAAAAAAAAAAAAAABbQ29udGVudF9UeXBlc10ueG1sUEsBAi0A&#10;FAAGAAgAAAAhADj9If/WAAAAlAEAAAsAAAAAAAAAAAAAAAAALwEAAF9yZWxzLy5yZWxzUEsBAi0A&#10;FAAGAAgAAAAhANCDmlBWAgAAsQQAAA4AAAAAAAAAAAAAAAAALgIAAGRycy9lMm9Eb2MueG1sUEsB&#10;Ai0AFAAGAAgAAAAhAGHXsvrfAAAACgEAAA8AAAAAAAAAAAAAAAAAsAQAAGRycy9kb3ducmV2Lnht&#10;bFBLBQYAAAAABAAEAPMAAAC8BQAAAAA=&#10;\" fillcolor=\"white [3201]\" strokeweight=\".5pt\">\n", text_box_id);
+    extract_docx_char_append_string(alloc, content, "            <v:textbox>\n");
+    extract_docx_char_append_string(alloc, content, "              <w:txbxContent>");
+
+    for (p=paragraph_begin; p<paragraph_end; ++p) {
+        paragraph_t* paragraph = page->paragraphs[p];
+        if (extract_document_to_docx_content_paragraph(alloc, state, paragraph, content)) goto end;
+    }
+
+    extract_docx_char_append_string(alloc, content, "\n");
+    extract_docx_char_append_string(alloc, content, "\n");
+    extract_docx_char_append_string(alloc, content, "              </w:txbxContent>\n");
+    extract_docx_char_append_string(alloc, content, "            </v:textbox>\n");
+    extract_docx_char_append_string(alloc, content, "          </v:shape>\n");
+    extract_docx_char_append_string(alloc, content, "        </w:pict>\n");
+    extract_docx_char_append_string(alloc, content, "      </mc:Fallback>\n");
+    extract_docx_char_append_string(alloc, content, "    </mc:AlternateContent>\n");
+    extract_docx_char_append_string(alloc, content, "  </w:r>\n");
+    extract_docx_char_append_string(alloc, content, "</w:p>");
+    e = 0;
+    end:
+    return e;
+}
+
+
+int extract_document_to_docx_content(
+        extract_alloc_t*    alloc,
+        document_t*         document,
+        int                 spacing,
+        int                 rotation,
+        int                 images,
+        extract_astring_t*  content
+        )
+{
+    int ret = -1;
+    int text_box_id = 0;
+    int p;
+    
+    /* Write paragraphs into <content>. */
+    for (p=0; p<document->pages_num; ++p) {
+        page_t* page = document->pages[p];
+        int p;
+        content_state_t state;
+        state.font_name = NULL;
+        state.font_size = 0;
+        state.font_bold = 0;
+        state.font_italic = 0;
+        state.ctm_prev = NULL;
+        
+        for (p=0; p<page->paragraphs_num; ++p) {
+            paragraph_t* paragraph = page->paragraphs[p];
+            const matrix_t* ctm = &paragraph->lines[0]->spans[0]->ctm;
+            double rotate = atan2(ctm->b, ctm->a);
+            
+            if (spacing
+                    && state.ctm_prev
+                    && paragraph->lines_num
+                    && paragraph->lines[0]->spans_num
+                    && matrix_cmp4(
+                            state.ctm_prev,
+                            &paragraph->lines[0]->spans[0]->ctm
+                            )
+                    ) {
+                /* Extra vertical space between paragraphs that were at
+                different angles in the original document. */
+                if (extract_docx_paragraph_empty(alloc, content)) goto end;
+            }
+
+            if (spacing) {
+                /* Extra vertical space between paragraphs. */
+                if (extract_docx_paragraph_empty(alloc, content)) goto end;
+            }
+            
+            if (rotation && rotate != 0) {
+            
+                /* Find extent of paragraphs with this same rotation. extent
+                will contain max width and max height of paragraphs, in units
+                before application of ctm, i.e. before rotation. */
+                point_t extent = {0, 0};
+                int p0 = p;
+                int p1;
+                
+                outf("rotate=%.2frad=%.1fdeg ctm: ef=(%f %f) abcd=(%f %f %f %f)",
+                        rotate, rotate * 180 / pi,
+                        ctm->e,
+                        ctm->f,
+                        ctm->a,
+                        ctm->b,
+                        ctm->c,
+                        ctm->d
+                        );
+                
+                {
+                    /* We assume that first span is at origin of text
+                    block. This assumes left-to-right text. */
+                    double rotate0 = rotate;
+                    const matrix_t* ctm0 = ctm;
+                    point_t origin = {
+                            paragraph->lines[0]->spans[0]->chars[0].x,
+                            paragraph->lines[0]->spans[0]->chars[0].y
+                            };
+                    matrix_t ctm_inverse = {1, 0, 0, 1, 0, 0};
+                    double ctm_det = ctm->a*ctm->d - ctm->b*ctm->c;
+                    if (ctm_det != 0) {
+                        ctm_inverse.a = +ctm->d / ctm_det;
+                        ctm_inverse.b = -ctm->b / ctm_det;
+                        ctm_inverse.c = -ctm->c / ctm_det;
+                        ctm_inverse.d = +ctm->a / ctm_det;
+                    }
+                    else {
+                        outf("cannot invert ctm=(%f %f %f %f)",
+                                ctm->a, ctm->b, ctm->c, ctm->d);
+                    }
+
+                    for (p=p0; p<page->paragraphs_num; ++p) {
+                        paragraph = page->paragraphs[p];
+                        ctm = &paragraph->lines[0]->spans[0]->ctm;
+                        rotate = atan2(ctm->b, ctm->a);
+                        if (rotate != rotate0) {
+                            break;
+                        }
+
+                        /* Update <extent>. */
+                        {
+                            int l;
+                            for (l=0; l<paragraph->lines_num; ++l) {
+                                line_t* line = paragraph->lines[l];
+                                span_t* span = line_span_last(line);
+                                char_t* char_ = span_char_last(span);
+                                double adv = char_->adv * matrix_expansion(span->trm);
+                                double x = char_->x + adv * cos(rotate);
+                                double y = char_->y + adv * sin(rotate);
+
+                                double dx = x - origin.x;
+                                double dy = y - origin.y;
+
+                                /* Position relative to origin and before box rotation. */
+                                double xx = ctm_inverse.a * dx + ctm_inverse.b * dy;
+                                double yy = ctm_inverse.c * dx + ctm_inverse.d * dy;
+                                yy = -yy;
+                                if (xx > extent.x) extent.x = xx;
+                                if (yy > extent.y) extent.y = yy;
+                                if (0) outf("rotate=%f p=%i: origin=(%f %f) xy=(%f %f) dxy=(%f %f) xxyy=(%f %f) span: %s",
+                                        rotate, p, origin.x, origin.y, x, y, dx, dy, xx, yy, span_string(alloc, span));
+                            }
+                        }
+                    }
+                    p1 = p;
+                    rotate = rotate0;
+                    ctm = ctm0;
+                    outf("rotate=%f p0=%i p1=%i. extent is: (%f %f)",
+                            rotate, p0, p1, extent.x, extent.y);
+                }
+                
+                /* Paragraphs p0..p1-1 have same rotation. We output them into
+                a single rotated text box. */
+                
+                /* We need unique id for text box. */
+                text_box_id += 1;
+                
+                {
+                    /* Angles are in units of 1/60,000 degree. */
+                    int rot = (int) (rotate * 180 / pi * 60000);
+
+                    /* <wp:anchor distT=\.. etc are in EMU - 1/360,000 of a cm.
+                    relativeHeight is z-ordering. (wp:positionV:wp:posOffset,
+                    wp:positionV:wp:posOffset) is position of origin of box in
+                    EMU.
+
+                    The box rotates about its centre but we want to rotate
+                    about the origin (top-left). So we correct the position of
+                    box by subtracting the vector that the top-left moves when
+                    rotated by angle <rotate> about the middle. */
+                    double point_to_emu = 12700;    /* https://en.wikipedia.org/wiki/Office_Open_XML_file_formats#DrawingML */
+                    int x = (int) (ctm->e * point_to_emu);
+                    int y = (int) (ctm->f * point_to_emu);
+                    int w = (int) (extent.x * point_to_emu);
+                    int h = (int) (extent.y * point_to_emu);
+                    int dx;
+                    int dy;
+
+                    if (0) outf("rotate: %f rad, %f deg. rot=%i", rotate, rotate*180/pi, rot);
+
+                    h *= 2;
+                    /* We can't predict how much space Word will actually
+                    require for the rotated text, so make the box have the
+                    original width but allow text to take extra vertical
+                    space. There doesn't seem to be a way to make the text box
+                    auto-grow to contain the text. */
+
+                    dx = (int) ((1-cos(rotate)) * w / 2.0 + sin(rotate) * h / 2.0);
+                    dy = (int) ((cos(rotate)-1) * h / 2.0 + sin(rotate) * w / 2.0);
+                    outf("ctm->e,f=%f,%f rotate=%f => x,y=%ik %ik dx,dy=%ik %ik",
+                            ctm->e,
+                            ctm->f,
+                            rotate * 180/pi,
+                            x/1000,
+                            y/1000,
+                            dx/1000,
+                            dy/1000
+                            );
+                    x -= dx;
+                    y -= -dy;
+
+                    if (extract_document_output_rotated_paragraphs(alloc, page, p0, p1, rot, x, y, w, h, text_box_id, content, &state)) goto end;
+                }
+                p = p1 - 1;
+                //p = page->paragraphs_num - 1;
+            }
+            else {
+                if (extract_document_to_docx_content_paragraph(alloc, &state, paragraph, content)) goto end;
+            }
+        
+        }
+        
+        if (images) {
+            int i;
+            for (i=0; i<page->images_num; ++i) {
+                extract_document_append_image(alloc, content, &page->images[i]);
+            }
+        }
+    }
+    ret = 0;
+
+    end:
+
+    return ret;
+}
+
+
+
+static int systemf(extract_alloc_t* alloc, const char* format, ...)
+/* Like system() but takes printf-style format and args. Also, if we return +ve
+we set errno to EIO. */
+{
+    int e;
+    char* command;
+    va_list va;
+    va_start(va, format);
+    e = extract_vasprintf(alloc, &command, format, va);
+    va_end(va);
+    if (e < 0) return e;
+    outf("running: %s", command);
+    e = system(command);
+    extract_free(alloc, &command);
+    if (e > 0) {
+        errno = EIO;
+    }
+    return e;
+}
+
+static int read_all(extract_alloc_t* alloc, FILE* in, char** o_out)
+/* Reads until eof into zero-terminated malloc'd buffer. */
+{
+    size_t  len = 0;
+    size_t  delta = 128;
+    for(;;) {
+        size_t n;
+        if (extract_realloc2(alloc, o_out, len, len + delta + 1)) {
+            extract_free(alloc, o_out);
+            return -1;
+        }
+        n = fread(*o_out + len, 1 /*size*/, delta /*nmemb*/, in);
+        len += n;
+        if (feof(in)) {
+            (*o_out)[len] = 0;
+            return 0;
+        }
+        if (ferror(in)) {
+            /* It's weird that fread() and ferror() don't set errno. */
+            errno = EIO;
+            extract_free(alloc, o_out);
+            return -1;
+        }
+    }
+}
+
+static int read_all_path(extract_alloc_t* alloc, const char* path, char** o_text)
+/* Reads entire file into zero-terminated malloc'd buffer. */
+{
+    int e = -1;
+    FILE* f = NULL;
+    f = fopen(path, "rb");
+    if (!f) goto end;
+    if (read_all(alloc, f, o_text)) goto end;
+    e = 0;
+    end:
+    if (f) fclose(f);
+    if (e) extract_free(alloc, &o_text);
+    return e;
+}
+
+static int write_all(const void* data, size_t data_size, const char* path)
+{
+    int e = -1;
+    FILE* f = fopen(path, "w");
+    if (!f) goto end;
+    if (fwrite(data, data_size, 1 /*nmemb*/, f) != 1) goto end;
+    e = 0;
+    end:
+    if (f) fclose(f);
+    return e;
+}
+
+static int extract_docx_content_insert(
+        extract_alloc_t*    alloc,
+        const char*         original,
+        const char*         mid_begin_name,
+        const char*         mid_end_name,
+        extract_astring_t*  contentss,
+        int                 contentss_num,
+        char**              o_out
+        )
+/* Creates a string consisting of <original> with all strings in <contentss>
+inserted into <original>'s <mid_begin_name>...<mid_end_name> region, and
+appends this string to *o_out. */
+{
+    int e = -1;
+    const char* mid_begin;
+    const char* mid_end;
+    extract_astring_t   out;
+    extract_astring_init(&out);
+    
+    mid_begin = strstr(original, mid_begin_name);
+    if (!mid_begin) {
+        outf("error: could not find '%s' in docx content",
+                mid_begin_name);
+        errno = ESRCH;
+        goto end;
+    }
+    mid_begin += strlen(mid_begin_name);
+    
+    mid_end = strstr(mid_begin, mid_end_name);
+    if (!mid_end) {
+        outf("error: could not find '%s' in docx content",
+                mid_end_name);
+        errno = ESRCH;
+        goto end;
+    }
+
+    if (extract_astring_catl(alloc, &out, original, mid_begin - original)) goto end;
+    {
+        int i;
+        for (i=0; i<contentss_num; ++i) {
+            if (extract_astring_catl(alloc, &out, contentss[i].chars, contentss[i].chars_num)) goto end;
+        }
+    }
+    if (extract_astring_cat(alloc, &out, mid_end)) goto end;
+    
+    *o_out = out.chars;
+    out.chars = NULL;
+    e = 0;
+    
+    end:
+    if (e) {
+        extract_astring_free(alloc, &out);
+        *o_out = NULL;
+    }
+    return e;
+}
+
+static int s_find_mid(const char* text, const char* begin, const char* end, const char** o_begin, const char** o_end)
+/* Sets *o_begin to end of first occurrence of <begin> in <text>, and *o_end to
+beginning of first occurtence of <end> in <text>. */
+{
+    *o_begin = strstr(text, begin);
+    if (!*o_begin) goto fail;
+    *o_begin += strlen(begin);
+    *o_end = strstr(*o_begin, end);
+    if (!*o_end) goto fail;
+    return 0;
+    fail:
+    errno = ESRCH;
+    return -1;
+}
+
+
+int extract_docx_content_item(
+        extract_alloc_t*    alloc,
+        extract_astring_t*  contentss,
+        int                 contentss_num,
+        images_t*           images,
+        const char*         name,
+        const char*         text,
+        char**              text2
+        )
+{
+    int e = -1;
+    extract_astring_t   temp;
+    extract_astring_init(&temp);
+    *text2 = NULL;
+    
+    if (0)
+    {}
+    else if (!strcmp(name, "[Content_Types].xml")) {
+        /* Add information about all image types that we are going to use. */
+        const char* begin;
+        const char* end;
+        const char* insert;
+        int it;
+        extract_astring_free(alloc, &temp);
+        outf("text: %s", text);
+        if (s_find_mid(text, "<Types ", "</Types>", &begin, &end)) goto end;
+
+        insert = begin;
+        insert = strchr(insert, '>');
+        assert(insert);
+        insert += 1;
+
+        if (extract_astring_catl(alloc, &temp, text, insert - text)) goto end;
+        outf("images->imagetypes_num=%i", images->imagetypes_num);
+        for (it=0; it<images->imagetypes_num; ++it) {
+            const char* imagetype = images->imagetypes[it];
+            if (extract_astring_cat(alloc, &temp, "<Default Extension=\"")) goto end;
+            if (extract_astring_cat(alloc, &temp, imagetype)) goto end;
+            if (extract_astring_cat(alloc, &temp, "\" ContentType=\"image/")) goto end;
+            if (extract_astring_cat(alloc, &temp, imagetype)) goto end;
+            if (extract_astring_cat(alloc, &temp, "\"/>")) goto end;
+        }
+        if (extract_astring_cat(alloc, &temp, insert)) goto end;
+        *text2 = temp.chars;
+        extract_astring_init(&temp);
+    }
+    else if (!strcmp(name, "word/_rels/document.xml.rels")) {
+        /* Add relationships between image ids and image names within docx
+        archive. */
+        const char* begin;
+        const char* end;
+        int         j;
+        extract_astring_free(alloc, &temp);
+        if (s_find_mid(text, "<Relationships", "</Relationships>", &begin, &end)) goto end;
+        if (extract_astring_catl(alloc, &temp, text, end - text)) goto end;
+        outf("images.images_num=%i", images->images_num);
+        for (j=0; j<images->images_num; ++j) {
+            image_t* image = &images->images[j];
+            if (extract_astring_cat(alloc, &temp, "<Relationship Id=\"")) goto end;
+            if (extract_astring_cat(alloc, &temp, image->id)) goto end;
+            if (extract_astring_cat(alloc, &temp, "\" Type=\"http://schemas.openxmlformats.org/officeDocument/2006/relationships/image\" Target=\"media/")) goto end;
+            if (extract_astring_cat(alloc, &temp, image->name)) goto end;
+            if (extract_astring_cat(alloc, &temp, "\"/>")) goto end;
+        }
+        if (extract_astring_cat(alloc, &temp, end)) goto end;
+        *text2 = temp.chars;
+        extract_astring_init(&temp);
+    }
+    else if (!strcmp(name, "word/document.xml")) {
+        /* Insert paragraphs content. */
+        if (extract_docx_content_insert(
+                alloc,
+                text,
+                "<w:body>",
+                "</w:body>",
+                contentss,
+                contentss_num,
+                text2
+                )) goto end;
+    }
+    else {
+        *text2 = NULL;
+    }
+    e = 0;
+    end:
+    if (e) {
+        /* We might have set <text2> to new content. */
+        extract_free(alloc, text2);
+        /* We might have used <temp> as a temporary buffer. */
+        extract_astring_free(alloc, &temp);
+    }
+    extract_astring_init(&temp);
+    return e;
+}
+
+        
+
+static int check_path_shell_safe(const char* path)
+/* Returns -1 with errno=EINVAL if <path> contains sequences that could make it
+unsafe in shell commands. */
+{
+    if (0
+            || strstr(path, "..")
+            || strchr(path, '\'')
+            || strchr(path, '"')
+            || strchr(path, ' ')
+            ) {
+        errno = EINVAL;
+        return -1;
+    }
+    return 0;
+}
+
+static int remove_directory(extract_alloc_t* alloc, const char* path)
+{
+    if (check_path_shell_safe(path)) {
+        outf("path_out is unsafe: %s", path);
+        return -1;
+    }
+    return systemf(alloc, "rm -r '%s'", path);
+}
+
+#ifdef _WIN32
+#include <direct.h>
+static int s_mkdir(const char* path, int mode)
+{
+    (void) mode;
+    return _mkdir(path);
+}
+#else
+static int s_mkdir(const char* path, int mode)
+{
+    return mkdir(path, mode);
+}
+#endif
+
+
+int extract_docx_write_template(
+        extract_alloc_t*    alloc,
+        extract_astring_t*  contentss,
+        int                 contentss_num,
+        images_t*           images,
+        const char*         path_template,
+        const char*         path_out,
+        int                 preserve_dir
+        )
+{
+    int     e = -1;
+    int     i;
+    char*   path_tempdir = NULL;
+    FILE*   f = NULL;
+    char*   path = NULL;
+    char*   text = NULL;
+    char*   text2 = NULL;
+
+    assert(path_out);
+    assert(path_template);
+    
+    if (check_path_shell_safe(path_out)) {
+        outf("path_out is unsafe: %s", path_out);
+        goto end;
+    }
+
+    outf("images->images_num=%i", images->images_num);
+    if (extract_asprintf(alloc, &path_tempdir, "%s.dir", path_out) < 0) goto end;
+    if (systemf(alloc, "rm -r '%s' 2>/dev/null", path_tempdir) < 0) goto end;
+
+    if (s_mkdir(path_tempdir, 0777)) {
+        outf("Failed to create directory: %s", path_tempdir);
+        goto end;
+    }
+
+    outf("Unzipping template document '%s' to tempdir: %s",
+            path_template, path_tempdir);
+    e = systemf(alloc, "unzip -q -d '%s' '%s'", path_tempdir, path_template);
+    if (e) {
+        outf("Failed to unzip %s into %s",
+                path_template, path_tempdir);
+        goto end;
+    }
+
+    /* Might be nice to iterate through all items in path_tempdir, but for now
+    we look at just the items that we know extract_docx_content_item() will
+    modify. */
+    
+    {
+        const char* names[] = {
+                "word/document.xml",
+                "[Content_Types].xml",
+                "word/_rels/document.xml.rels",
+                };
+        int names_num = sizeof(names) / sizeof(names[0]);
+        for (i=0; i<names_num; ++i) {
+            const char* name = names[i];
+            extract_free(alloc, &path);
+            extract_free(alloc, &text);
+            extract_free(alloc, &text2);
+            if (extract_asprintf(alloc, &path, "%s/%s", path_tempdir, name) < 0) goto end;
+            if (read_all_path(alloc, path, &text)) goto end;
+            
+            if (extract_docx_content_item(
+                    alloc,
+                    contentss,
+                    contentss_num,
+                    images,
+                    name,
+                    text,
+                    &text2
+                    )) goto end;
+            {
+                const char* text3 = (text2) ? text2 : text;
+                if (write_all(text3, strlen(text3), path)) goto end;
+            }
+        }
+    }
+
+    /* Copy images into <path_tempdir>/media/. */
+    extract_free(alloc, &path);
+    if (extract_asprintf(alloc, &path, "%s/word/media", path_tempdir) < 0) goto end;
+    if (s_mkdir(path, 0777)) goto end;
+    
+    for (i=0; i<images->images_num; ++i) {
+        image_t* image = &images->images[i];
+        extract_free(alloc, &path);
+        if (extract_asprintf(alloc, &path, "%s/word/media/%s", path_tempdir, image->name) < 0) goto end;
+        if (write_all(image->data, image->data_size, path)) goto end;
+    }
+    
+    outf("Zipping tempdir to create %s", path_out);
+    {
+        const char* path_out_leaf = strrchr(path_out, '/');
+        if (!path_out_leaf) path_out_leaf = path_out;
+        e = systemf(alloc, "cd '%s' && zip -q -r -D '../%s' .", path_tempdir, path_out_leaf);
+        if (e) {
+            outf("Zip command failed to convert '%s' directory into output file: %s",
+                    path_tempdir, path_out);
+            goto end;
+        }
+    }
+
+    if (!preserve_dir) {
+        if (remove_directory(alloc, path_tempdir)) goto end;
+    }
+
+    e = 0;
+
+    end:
+    outf("e=%i", e);
+    extract_free(alloc, &path_tempdir);
+    extract_free(alloc, &path);
+    extract_free(alloc, &text);
+    extract_free(alloc, &text2);
+    if (f)  fclose(f);
+
+    if (e) {
+        outf("Failed to create %s", path_out);
+    }
+    return e;
+}