Handle empty blog case in README recipe.

[redo-blog] / README.md

diff --git a/README.md b/README.md
index e4d63fa5f6c218e056a13cf37c634181b37a3400..d9e0f2c0e73618aa0c4cb783eb6f8278f9a82121 100644 (file)
--- a/README.md
+++ b/README.md
@@ -1,7 +1,8 @@
 redo-blog
 =========
 
-small blog system using the redo build system
+small blog system using the redo build system, with blog article files written
+in (pandoc) Markdown or ReStructured Text.
 
 dependencies
 ------------
@@ -28,13 +29,16 @@ You can then enter the directory and run redo there. This will generate article
 These files will be linked to symbolically in a directory ./public/.
 
 Some metadata files will also be generated below ./metadata/: For each article,
-there will be generated a .uuid and a .intermediate file; furthermore, files for
-data used in ./feed.xml and ./index.html will be built there and can be edited
-to customize the blog – namely the files url, author, uuid, title, index.tmpl,
-index_snippet.tmpl, article.tmpl.
+there will be generated a .automatic_metadata (to contain an article's UUID,
+checksum, and creation/modification dates) and a .intermediate file (to contain
+pandoc-formatted article content like title and body); furthermore, files for
+data used in ./feed.xml and ./index.html will, if non-existant, be built there
+and can be edited to customize the blog – namely the files url, author, title,
+index.tmpl, index_snippet.tmpl, article.tmpl. A blog-specific UUID and creation
+date is stored in ./metadata/automatic_metadata
 
-recipe to use server-side with git
-----------------------------------
+recipe to remotely manage a redo blog with git
+----------------------------------------------
 
 On your server, install the dependencies listed above. Then set up a repository
 for your blog files. Let's assume we want it to sit in our home directory and be
@@ -45,10 +49,22 @@ called `blog`:
     git init --bare blog.git
     cat << EOF > blog.git/hooks/post-update
     #!/bin/sh
-    BLOGDIR=~/blog
-    GIT_WORK_TREE=$BLOGDIR git checkout -f
-    cd \$BLOGDIR
+    blog_dir=~/blog
+    export GIT_DIR=\$(pwd)
+    export GIT_WORK_TREE="\$blog_dir"
+    git checkout -f
+    cd "\$GIT_WORK_TREE"
     redo
+    git add metadata/author metadata/url metadata/title metadata/*.tmpl metadata/automatic_metadata
+    count=\$(ls -1 metadata/*.automatic_metadata 2>/dev/null | wc -l)
+    if [ "\$count" != 0 ]; then
+      git add metadata/*.automatic_metadata
+    fi
+    status=\$(git status -s)
+    n_updates=\$(printf "$status" | grep -vE '^\?\?' | wc -l)
+    if [ "\$n_updates" -gt 0 ]; then
+      git commit -a -m 'Update metadata'
+    fi
     EOF
     chmod a+x blog.git/hooks/post-update
 
@@ -64,7 +80,8 @@ public web content to sit:
 
     ln -s ~/blog/public /var/www/html/blog
 
-Client-side, do this (obviously, replace server and username):
+Client-side, do this (you obviously need to customize this code; at least
+replace the username `user` and the server name `example.org`):
 
     cd ~
     git init blog
@@ -76,11 +93,19 @@ Client-side, do this (obviously, replace server and username):
     git commit -m 'set up blog metadata'
     git push origin master
 
-bugs
-----
+If successful, the git hook will furthermore commit some ~/blog/metadata/ files
+generated by redo, that can be pulled into the client-side local repository:
+
+    git pull origin master
+
+bugs and peculiarities
+----------------------
 
 Don't create a index.rst or index.md in the redo-managed directory, that will
 break things.
 
-Running redo without a single blog entry source file (.md or .rst) in the
-directory will also break things.
+The article title is derived in .md files from a first line prefixed with `%`,
+while all other headings are treated as sub-headings. In .rst files, the title
+is derived from a heading that must be at the top of the document, and be of an
+adornment style (such as underlining with `=`) that must be used only once in
+it.