Make some fields in the templates required

It's annoying that we keep needing two URLs for everything. One to
server the UI and one to actually do it: edit/save (with page name in
the URL), add/append (that's a sketchy pair), upload/save (save again,
but this time without a trailing slash, ugh).

.gitignore

@@ -1,2 +1,3 @@
 /oddmu
 test.md
+/testdata/

Makefile

@@ -25,4 +25,4 @@ test:
 upload:
 	go build
 	rsync --itemize-changes --archive oddmu oddmu.service *.html README.md sibirocobombus.root:/home/oddmu/
-	ssh sibirocobombus.root "systemctl restart oddmu"
+	ssh sibirocobombus.root "systemctl restart oddmu; systemctl restart alex"

README.md

@@ -1,40 +1,136 @@
 # Oddµ: A minimal wiki
 
 This program runs a wiki. It serves all the Markdown files (ending in
-`.md`) into web pages and allows you to edit them.
+`.md`) into web pages and allows you to edit them. If your files don't
+provide their own title (`# title`), the file name (without `.md`) is
+used for the title. Subdirectories are created as necessary.
 
-This is a minimal wiki. There is no version history. It probably makes
-sense to only use it as one person or in very small groups.
+This is a minimal wiki. There is no version history. It's well suited
+as a *secondary* medium: collaboration and conversation happens
+elsewhere, in chat, on social media. The wiki serves as the text
+repository that results from these discussions.
 
-This wiki only uses Markdown. There is no additional wiki markup, most
-importantly double square brackets are not a link. If you're used to
+The wiki lists no recent changes. The expectation is that the people
+that care were involved in the discussions beforehand.
+
+The wiki also produces no feed. The assumption is that announcements
+are made on social media: blogs, news aggregators, discussion forums,
+the fediverse, but humans. There is no need for bots.
+
+As you'll see below, the idea is that the webserver handles as many
+tasks as possible. It logs requests, does rate limiting, handles
+encryption, gets the certificates, and so on. The web server acts as a
+reverse proxy and the wiki ends up being a content management system
+with almost no structure – or endless malleability, depending on your
+point of view.
+
+And last but not least: µ is the letter mu, so Oddµ is usually written
+Oddmu. 🙃
+
+## Markdown
+
+This wiki uses a [Markdown
+library](https://github.com/gomarkdown/markdown) to generate the web
+pages from Markdown. There is no additional wiki markup. Most
+importantly, double square brackets are not a link. If you're used to
 that, it'll be strange as you need to repeat the name: `[like
 this](like this)`.
 
-If your files don't provide their own title (`# title`), the file name
-is used for the title.
+The Markdown processor comes with a few extensions, some of which are
+enable by default:
 
-µ is the letter mu, so Oddµ is usually written Oddmu. 🙃
+* emphasis markers inside words are ignored
+* tables are supported
+* fenced code blocks are supported
+* autolinking of "naked" URLs are supported
+* strikethrough using two tildes is supported (`~~like this~~`)
+* it is strict about prefix heading rules
+* you can specify an id for headings (`{#id}`)
+* trailing backslashes turn into line breaks
+* definition lists are supported
+* MathJax is supported (but needs a separte setup)
+
+A table with footers and a columnspan:
+
+```text
+Name    | Age
+--------|------
+Bob     ||
+Alice   | 23
+========|======
+Total   | 23
+```
+
+A definition list:
+
+```text
+Cat
+: Fluffy animal everyone likes
+
+Internet
+: Vector of transmission for pictures of cats
+```
+
+There is another extension made: hashtags link to searches for the
+hashtag. Hashtags are separate from titles because there is no space
+after the hash. Use the underscore to use hashtags consisting of
+multiple words.
+
+```
+# Title
+
+Text
+
+#Tag #Another_Tag
+```
 
 ## Templates
 
-Feel free to change the templates `view.html` and `edit.html` and
-restart the server. Modifying the styles in the templates would be a
-good start to get a feel for it.
+The template files are the HTML files in the working directory:
+`add.html`, `edit.html`, `search.html`, `upload.html` and `view.html`.
+Feel free to change the templates and restart the server. The first
+change you should make is to replace the email address in `view.html`.
+😄
+
+See [Structuring the web
+with HTML](https://developer.mozilla.org/en-US/docs/Learn/HTML) to
+learn more about HTML.
+
+Modifying the styles in the templates would be another good start to
+get a feel for it. See [Learn to style HTML using
+CSS](https://developer.mozilla.org/en-US/docs/Learn/CSS) to learn more
+about style sheets.
 
 The templates can refer to the following properties of a page:
 
 `{{.Title}}` is the page title. If the page doesn't provide its own
 title, the page name is used.
 
-`{{.Name}}` is the page name. The page name doesn't include the `.md`
-extension.
+`{{.Name}}` is the page name, escaped for use in URLs. More
+specifically, it is URI escaped except for the slashes. The page name
+doesn't include the `.md` extension.
 
 `{{.Html}}` is the rendered Markdown, as HTML.
 
 `{{printf "%s" .Body}}` is the Markdown, as a string (the data itself
 is a byte array and that's why we need to call `printf`).
 
+For the `search.html` template only:
+
+`{{.Results}}` indicates if there were any search results.
+
+`{{.Items}}` is an array of pages, each containing a search result. A
+search result is a page (with the properties seen above). Thus, to
+refer to them, you need to use a `{{range .Items}}` … `{{end}}`
+construct.
+
+For search results, `{{.Html}}` is the rendered Markdown of a page
+summary, as HTML.
+
+`{{.Score}}` is a numerical score for search results.
+
+The `upload.html` template cannot refer to anything.
+
 When calling the `save` action, the page name is take from the URL and
 the page content is taken from the `body` form parameter. To
 illustrate, here's how to edit a page using `curl`:
@@ -44,6 +140,25 @@ curl --form body="Did you bring a towel?" \
   http://localhost:8080/save/welcome
 ```
 
+The wiki uses the standard
+[html/template](https://pkg.go.dev/html/template) library to do this.
+There's more information on writing templates in the documentation for
+the [text/template](https://pkg.go.dev/text/template) library.
+
+## Non-English hyphenation
+
+Automatic hyphenation by the browser requires two things: The style
+sheet must indicate `hyphen: auto` for an HTML element such as `body`,
+and that element must have a `lang` set (usually a two letter language
+code such as `de` for German). This happens in the template files,
+such as `view.html` and `search.html`.
+
+Oddmu uses the [lingua](github.com/pemistahl/lingua-go) library to
+detect languages. If you know that you're only going to use a small
+number of languages – or just a single language! – you can set the
+environment variable ODDMU_LANGUAGES to a comma-separated list of ISO
+639-1 codes, e.g. "en" or "en,de,fr,pt".
+
 ## Building
 
 ```sh
@@ -69,6 +184,9 @@ If you ran it in the source directory, try
 http://localhost:8080/view/README – this serves the README file you're
 currently reading.
 
+You can change the port by setting the ODDMU_PORT environment
+variable.
+
 ## Deploying it using systemd
 
 As root, on your server:
@@ -77,14 +195,17 @@ As root, on your server:
 adduser --system --home /home/oddmu oddmu
 ```
 
-Copy all the files into `/home/oddmu` to your server: `oddmu`, `oddmu.service`, `view.html` and `edit.html`.
+Copy all the files into `/home/oddmu` to your server: `oddmu`,
+`oddmu.service`, and all the template files ending in `.html`.
 
-Edit the `oddmu.service` file. These are the three lines you most likely have to take care of:
+Edit the `oddmu.service` file. These are the three lines you most
+likely have to take care of:
 
 ```
 ExecStart=/home/oddmu/oddmu
 WorkingDirectory=/home/oddmu
 Environment="ODDMU_PORT=8080"
+Environment="ODDMU_LANGUAGES=en,de"
 ```
 
 Install the service file and enable it:
@@ -132,27 +253,21 @@ MDCertificateAgreement accepted
     ServerAdmin alex@alexschroeder.ch
     ServerName transjovian.org
     SSLEngine on
-
-    RewriteEngine on
-    RewriteRule ^/$ http://%{HTTP_HOST}:8080/view/index [redirect]
-    RewriteRule ^/(view|edit|save|search)/(.*) http://%{HTTP_HOST}:8080/$1/$2 [proxy]
+    ProxyPassMatch ^/(search|upload|save|(view|edit|save|add|append)/(.*))?$ http://localhost:8080/$1
 </VirtualHost>
 ```
 
 First, it manages the domain, getting the necessary certificates. It
 redirects regular HTTP traffic from port 80 to port 443. It turns on
-the SSL engine for port 443. It redirects `/` to `/view/index` and any
-path that starts with `/view/`, `/edit/`, `/save/` or `/search/` is
-proxied to port 8080 where the Oddmu program can handle it.
+the SSL engine for port 443. It proxies the requests for the wiki to
+port 8080.
 
 Thus, this is what happens:
 
-* The user tells the browser to visit `http://transjovian.org` (on port 80)
-* Apache redirects this to `http://transjovian.org/` by default (still on port 80)
-* Our first virtual host redirects this to `https://transjovian.org/` (encrypted, on port 443)
-* Our second virtual host redirects this to `https://transjovian.org/wiki/view/index` (still on port 443)
-* This is proxied to `http://transjovian.org:8080/view/index` (no on port 8080, without encryption)
-* The wiki converts `index.md` to HTML, adds it to the template, and serves it.
+* The user tells the browser to visit `transjovian.org`
+* The browser sends a request for `http://transjovian.org` (on port 80)
+* Apache redirects this to `https://transjovian.org/` by default (now on port 443)
+* This is proxied to `http://transjovian.org:8080/` (no encryption, on port 8080)
 
 Restart the server, gracefully:
 
@@ -160,6 +275,11 @@ Restart the server, gracefully:
 apachectl graceful
 ```
 
+To serve both HTTP and HTTPS, don't redirect from the first virtual
+host to the second – instead just proxy to the wiki like you did for
+the second virtual host: use a copy of the `ProxyPassMatch` directive
+instead of `RewriteEngine on` and `RewriteRule`.
+
 ## Access
 
 Access control is not part of the wiki. By default, the wiki is
@@ -190,12 +310,12 @@ To delete remove a user:
 htpasswd -D .htpasswd berta
 ```
 
-Modify your site configuration and protect the `/edit/` and `/save/`
-URLs with a password by adding the following to your `<VirtualHost
-*:443>` section:
+Modify your site configuration and protect the `/edit/`, `/save/`,
+`/add/`, `/append/`, `/upload` and `/save` URLs with a password by
+adding the following to your `<VirtualHost *:443>` section:
 
 ```apache
-<LocationMatch "^/(edit|save)/">
+<LocationMatch "^/(upload|save|(edit|save|add|append)/(.*))$">
   AuthType Basic
   AuthName "Password Required"
   AuthUserFile /home/oddmu/.htpasswd
@@ -220,10 +340,9 @@ webserver can read (world readable file, world readable and executable
 directory). Populate it with files.
 
 Make sure that none of the static files look like the wiki paths
-`/view/`, `/edit/`, `/save/` or `/search/`.
-
-For example, create a file called `robots.txt` containing the
-following, tellin all robots that they're not welcome.
+`/view/`, `/edit/`, `/save/`, `/add/`, `/append/`, `/upload`, `/save`
+or `/search`. For example, create a file called `robots.txt`
+containing the following, tellin all robots that they're not welcome.
 
 ```text
 User-agent: *
@@ -236,62 +355,103 @@ and without needing a wiki page.
 [Wikipedia](https://en.wikipedia.org/wiki/Robot_exclusion_standard)
 has more information.
 
-## Customization (with recompilation)
+## Different logins for different access rights
 
-The Markdown parser can be customized and
-[extensions](https://pkg.go.dev/github.com/gomarkdown/markdown/parser#Extensions)
-can be added. There's an example in the
-[usage](https://github.com/gomarkdown/markdown#usage) section. You'll
-need to make changes to the `viewHandler` yourself.
+What if you have a site with various subdirectories and each
+subdirectory is for a different group of friends? You can set this up
+using your webserver. One way to do this is to require specific
+usernames (which must have a password in the password file mentioned
+above.
 
-### Render Gemtext
+This requires a valid login by the user "alex" or "berta":
 
-In a first approximation, Gemtext is valid Markdown except for the
-rocket links (`=>`). Here's how to modify the `loadPage` so that a
-`.gmi` file is loaded if no `.md` is found, and the rocket links are
-translated into Markdown:
-
-```go
-func loadPage(name string) (*Page, error) {
-	filename := name + ".md"
-	body, err := os.ReadFile(filename)
-	if err == nil {
-		return &Page{Title: name, Name: name, Body: body}, nil
-	}
-	filename = name + ".gmi"
-	body, err = os.ReadFile(filename)
-	if err == nil {
-		return &Page{Title: name, Name: name, Body: body}, nil
-	}
-	return nil, err
-}
+```apache
+<LocationMatch "^/(edit|save|add|append)/intetebi/">
+  Require user alex berta
+</LocationMatch>
 ```
 
-There is a small problem, however: By default, Markdown expects an
-empty line before a list begins. The following change to `renderHtml`
-uses the `NoEmptyLineBeforeBlock` extension for the parser:
+## Private wikis
 
-```go
-func (p* Page) renderHtml() {
-    // Here is where a new extension is added!
-	extensions := parser.CommonExtensions | parser.NoEmptyLineBeforeBlock
-	markdownParser := parser.NewWithExtensions(extensions)
-	maybeUnsafeHTML := markdown.ToHTML(p.Body, markdownParser, nil)
-	html := bluemonday.UGCPolicy().SanitizeBytes(maybeUnsafeHTML)
-	p.Html = template.HTML(html);
-}
+Based on the above, you can prevent people from reading the wiki, too.
+The `LocationMatch` must cover the `/view/` URLs. In order to protect
+*everything*, use a [Location directive](https://httpd.apache.org/docs/current/mod/core.html#location)
+that matches everything:
+
+```apache
+<Location />
+  AuthType Basic
+  AuthName "Password Required"
+  AuthUserFile /home/oddmu/.htpasswd
+  Require valid-user
+</Location>
 ```
 
+## Virtual hosting
+
+Virtual hosting in this context means that the program serves two
+different sites for two different domains from the same machine. Oddmu
+doesn't support that, but your webserver does. Therefore, start an
+Oddmu instance for every domain name, each listening on a different
+port. Then set up your web server such that ever domain acts as a
+reverse proxy to a different Oddmu instance.
+
+## Understanding search
+
+The index indexes trigrams. Each group of three characters is a
+trigram. A document with content "This is a test" is turned to lower
+case and indexed under the trigrams "thi", "his", "is ", "s i", " is",
+"is ", "s a", " a ", "a t", " te", "tes", "est".
+
+Each query is split into words and then processed the same way. A
+query with the words "this test" is turned to lower case and produces
+the trigrams "thi", "his", "tes", "est". This means that the word
+order is not considered when searching for documents.
+
+This also means that there is no stemming. Searching for "testing"
+won't find "This is a test" because there are no matches for the
+trigrams "sti", "tin", "ing".
+
+These trigrams are looked up in the index, resulting in the list of
+documents. Each document found is then scored. Each of the following
+increases the score by one point:
+
+- the entire phrase matches
+- a word matches
+- a word matches at the beginning of a word
+- a word matches at the end of a word
+- a word matches as a whole word
+
+A document with content "This is a test" when searched with the phrase
+"this test" therefore gets a score of 8: the entire phrase does not
+match but each word gets four points.
+
+Trigrams are sometimes strange: In a text containing the words "main"
+and "rail", a search for "mail" returns a match because the trigrams
+"mai" and "ail" are found. In this situation, the result has a score
+of 0.
+
 ## Limitations
 
 Page titles are filenames with `.md` appended. If your filesystem
-cannot handle it, it can't be a page title. Specifically, *no slashes*
-in filenames.
+cannot handle it, it can't be a page name.
 
 The pages are indexed as the server starts and the index is kept in
 memory. If you have a ton of pages, this surely wastes a lot of
 memory.
 
+Files may not end with a tilde (`~`) – these are backup files.
+
+You cannot edit uploaded files. If you upload a file called
+`hello.txt` and attempt to edit it by using `/edit/hello.txt` you will
+create a page with the name `hello.txt.md` instead.
+
+You cannot delete uploaded files via the web.
+
+## Bugs
+
+If you spot any, [contact](https://alexschroeder.ch/wiki/Contact) me.
+
 ## References
 
 [Writing Web Applications](https://golang.org/doc/articles/wiki/)

TODO.md

@@ -0,0 +1,6 @@
+Upload files should use path info so that we can use Apache to
+restrict access to directories.
+
+Automatically scale or process files.
+
+Post by Delta Chat? That is, allow certain encrypted emails to post.

add.html

@@ -0,0 +1,21 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <meta charset="utf-8">
+    <meta name="format-detection" content="telephone=no">
+    <meta name="viewport" content="width=device-width">
+    <title>Add to {{.Title}}</title>
+    <style>
+html { max-width: 70ch; padding: 2ch; margin: auto; color: #111; background: #ffe; }
+form, textarea { width: 100%; }
+    </style>
+  </head>
+  <body>
+    <h1>Adding to {{.Title}}</h1>
+    <form action="/append/{{.Name}}" method="POST">
+      <textarea name="body" rows="20" cols="80" placeholder="Text" autofocus required></textarea>
+      <p><input type="submit" value="Add">
+        <a href="/view/{{.Name}}"><button>Cancel</button></a></p>
+    </form>
+  </body>
+</html>

commands.go

@@ -0,0 +1,25 @@
+package main
+
+import (
+	"fmt"
+	"os"
+)
+
+func commands() {
+	if len(os.Args) == 3 && os.Args[1] == "html" {
+		p, err := loadPage(os.Args[2]);
+		if err != nil {
+			fmt.Println(err);
+		} else {
+			p.renderHtml();
+			fmt.Println(p.Html);
+		}
+	} else {
+		fmt.Printf("Unknown command: %v\n", os.Args[1:])
+		fmt.Print("Without any arguments, serves a wiki.\n")
+		fmt.Print(" Environment variable ODDMUSE_PORT controls the port.\n")
+		fmt.Print(" Environment variable ODDMUSE_LANGAUGES controls the languages detected.\n")
+		fmt.Print("html PAGENAME\n")
+		fmt.Print(" Print the HTML of the page.\n")
+	}
+}

concurrency_test.go

@@ -0,0 +1,14 @@
+package main
+
+import (
+	"github.com/stretchr/testify/assert"
+	"testing"
+)
+
+// Use go test -race to see whether this is a race condition.
+func TestLoadAndSearch(t *testing.T) {
+	go loadIndex()
+	q := "Oddµ"
+	pages := search(q)
+	assert.Zero(t, len(pages))
+}

edit.html

@@ -12,11 +12,12 @@ form, textarea { width: 100%; }
   </head>
   <body>
     <h1>Editing {{.Title}}</h1>
-
     <form action="/save/{{.Name}}" method="POST">
-      <div><textarea name="body" rows="20" cols="80">{{printf "%s" .Body}}</textarea></div>
+      <textarea name="body" rows="20" cols="80" placeholder="# Title
+
+Text" autofocus>{{printf "%s" .Body}}</textarea>
       <p><input type="submit" value="Save">
-        <a href="/view/{{.Name}}"><button>Cancel</button></a></p>
+        <a href="/view/{{.Name}}"><button type="button">Cancel</button></a></p>
     </form>
   </body>
 </html>

go.mod

@@ -4,12 +4,20 @@ go 1.21.0
 
 require (
 	github.com/dgryski/go-trigram v0.0.0-20160407183937-79ec494e1ad0
-	github.com/gomarkdown/markdown v0.0.0-20230716120725-531d2d74bc12
+	github.com/gomarkdown/markdown v0.0.0-20230912175223-14b07df9d538
 	github.com/microcosm-cc/bluemonday v1.0.25
+	github.com/pemistahl/lingua-go v1.4.0
+	github.com/stretchr/testify v1.8.4
 )
 
 require (
 	github.com/aymerick/douceur v0.2.0 // indirect
+	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/gorilla/css v1.0.0 // indirect
+	github.com/pmezard/go-difflib v1.0.0 // indirect
+	github.com/shopspring/decimal v1.3.1 // indirect
+	golang.org/x/exp v0.0.0-20221106115401-f9659909a136 // indirect
 	golang.org/x/net v0.12.0 // indirect
+	google.golang.org/protobuf v1.31.0 // indirect
+	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -1,12 +1,38 @@
 github.com/aymerick/douceur v0.2.0 h1:Mv+mAeH1Q+n9Fr+oyamOlAkUNPWPlA8PPGR0QAaYuPk=
 github.com/aymerick/douceur v0.2.0/go.mod h1:wlT5vV2O3h55X9m7iVYN0TBM0NH/MmbLnd30/FjWUq4=
+github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
+github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/dgryski/go-trigram v0.0.0-20160407183937-79ec494e1ad0 h1:b+7JSiBM+hnLQjP/lXztks5hnLt1PS46hktG9VOJgzo=
 github.com/dgryski/go-trigram v0.0.0-20160407183937-79ec494e1ad0/go.mod h1:qzKC/DpcxK67zaSHdCmIv3L9WJViHVinYXN2S7l3RM8=
+github.com/golang/protobuf v1.5.0/go.mod h1:FsONVRAS9T7sI+LIUmWTfcYkHO4aIWwzhcaSAoJOfIk=
 github.com/gomarkdown/markdown v0.0.0-20230716120725-531d2d74bc12 h1:uK3X/2mt4tbSGoHvbLBHUny7CKiuwUip3MArtukol4E=
 github.com/gomarkdown/markdown v0.0.0-20230716120725-531d2d74bc12/go.mod h1:JDGcbDT52eL4fju3sZ4TeHGsQwhG9nbDV21aMyhwPoA=
+github.com/gomarkdown/markdown v0.0.0-20230912175223-14b07df9d538 h1:ePDpFu7l0QUV46/9A7icfL2wvIOzTJLCWh4RO2NECzE=
+github.com/gomarkdown/markdown v0.0.0-20230912175223-14b07df9d538/go.mod h1:JDGcbDT52eL4fju3sZ4TeHGsQwhG9nbDV21aMyhwPoA=
+github.com/google/go-cmp v0.5.5/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
+github.com/google/go-cmp v0.5.8 h1:e6P7q2lk1O+qJJb4BtCQXlK8vWEO8V1ZeuEdJNOqZyg=
+github.com/google/go-cmp v0.5.8/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
 github.com/gorilla/css v1.0.0 h1:BQqNyPTi50JCFMTw/b67hByjMVXZRwGha6wxVGkeihY=
 github.com/gorilla/css v1.0.0/go.mod h1:Dn721qIggHpt4+EFCcTLTU/vk5ySda2ReITrtgBl60c=
 github.com/microcosm-cc/bluemonday v1.0.25 h1:4NEwSfiJ+Wva0VxN5B8OwMicaJvD8r9tlJWm9rtloEg=
 github.com/microcosm-cc/bluemonday v1.0.25/go.mod h1:ZIOjCQp1OrzBBPIJmfX4qDYFuhU02nx4bn030ixfHLE=
+github.com/pemistahl/lingua-go v1.4.0 h1:ifYhthrlW7iO4icdubwlduYnmwU37V1sbNrwhKBR4rM=
+github.com/pemistahl/lingua-go v1.4.0/go.mod h1:ECuM1Hp/3hvyh7k8aWSqNCPlTxLemFZsRjocUf3KgME=
+github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
+github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
+github.com/shopspring/decimal v1.3.1 h1:2Usl1nmF/WZucqkFZhnfFYxxxu8LG21F6nPQBE5gKV8=
+github.com/shopspring/decimal v1.3.1/go.mod h1:DKyhrW/HYNuLGql+MJL6WCR6knT2jwCFRcu2hWCYk4o=
+github.com/stretchr/testify v1.8.4 h1:CcVxjf3Q8PM0mHUKJCdn+eZZtm5yQwehR5yeSVQQcUk=
+github.com/stretchr/testify v1.8.4/go.mod h1:sz/lmYIOXD/1dqDmKjjqLyZ2RngseejIcXlSw2iwfAo=
+golang.org/x/exp v0.0.0-20221106115401-f9659909a136 h1:Fq7F/w7MAa1KJ5bt2aJ62ihqp9HDcRuyILskkpIAurw=
+golang.org/x/exp v0.0.0-20221106115401-f9659909a136/go.mod h1:CxIveKay+FTh1D0yPZemJVgC/95VzuuOLq5Qi4xnoYc=
 golang.org/x/net v0.12.0 h1:cfawfvKITfUsFCeJIHJrbSxpeu/E81khclypR0GVT50=
 golang.org/x/net v0.12.0/go.mod h1:zEVYFnQC7m/vmpQFELhcD1EWkZlX69l4oqgmer6hfKA=
+golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
+google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
+google.golang.org/protobuf v1.31.0 h1:g0LDEJHgrBl9N9r17Ru3sqWhkIx2NB67okBHPwC7hs8=
+google.golang.org/protobuf v1.31.0/go.mod h1:HV8QOd/L58Z+nl8r43ehVNZIU/HEI6OcFqwMG9pJV4I=
+gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
+gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
+gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
+gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

highlight.go

@@ -1,45 +1,13 @@
 package main
 
 import (
-	"strings"
 	"regexp"
 )
 
 // highlight splits the query string q into terms and highlights them
-// using the bold tag. Return the highlighted string and a score.
-func highlight (q string, s string) (string, int) {
-	c := 0
-	re, err := regexp.Compile("(?i)" + q)
-	if err == nil {
-		m := re.FindAllString(s, -1)
-		if m != nil {
-			// Score increases for each full match of q.
-			c += len(m)
-		}
-	}
-	for _, v := range strings.Split(q, " ") {
-		if len(v) == 0 {
-			continue
-		}
-		re, err := regexp.Compile(`(?is)(\pL?)(` + v + `)(\pL?)`)
-		if err != nil {
-			continue
-		}
-		r := make(map[string]string)
-		for _, m := range re.FindAllStringSubmatch(s, -1) {
-			// Term matched increases the score.
-			c++
-			// Terms matching at the beginning and
-			// end of words and matching entire
-			// words increase the score further.
-			if len(m[1]) == 0 { c++ }
-			if len(m[3]) == 0 { c++ }
-			if len(m[1]) == 0 && len(m[3]) == 0 { c++ }
-			r[m[2]] = "<b>" + m[2] + "</b>"
-		}
-		for old, new := range r {
-			s = strings.ReplaceAll(s, old, new)
-		}
-	}
-	return s, c
+// using the bold tag. Return the highlighted string.
+// This assumes that q already has all its meta characters quoted.
+func highlight(q string, re *regexp.Regexp, s string) string {
+	s = re.ReplaceAllString(s, "<b>$1</b>")
+	return s
 }

highlight_test.go

@@ -15,49 +15,29 @@ A wave of car noise hits me
 No birds to be heard.`
 
 	q := "window"
-	r, c := highlight(q, s)
+	re, _ := re(q)
+	r := highlight(q, re, s)
 	if r != h {
 		t.Logf("The highlighting is wrong in ｢%s｣", r)
 		t.Fail()
 	}
-	// Score:
-	// - q itself
-	// - the single token
-	// - the beginning of a word
-	if c != 3 {
-		t.Logf("%s score is %d", q, c)
-		t.Fail()
-	}
-	q = "windows"
-	_, c = highlight(q, s)
-	// Score:
-	// - q itself
-	// - the single token
-	// - the beginning of a word
-	// - the end of a word
-	// - the whole word
-	if c != 5 {
-		t.Logf("%s score is %d", q, c)
-		t.Fail()
-	}
-	q = "car noise"
-	_, c = highlight(q, s)
-	// Score:
-	// - car noise (+1)
-	// - car, with beginning, end, whole word (+4)
-	// - noise, with beginning, end, whole word (+4)
-	if c != 9 {
-		t.Logf("%s score is %d", q, c)
-		t.Fail()
-	}
-	q = "noise car"
-	_, c = highlight(q, s)
-	// Score:
-	// - the car token
-	// - the noise token
-	// - each with beginning, end and whole token (3 each)
-	if c != 8 {
-		t.Logf("%s score is %d", q, c)
+}
+
+func TestOverlap(t *testing.T) {
+
+	s := `Sit with me my love
+Kids shout and so do parents
+I hear the fountain`
+
+	h := `Sit with me my love
+Kids <b>shout</b> and so do parents
+I hear the fountain`
+
+	q := "shout out"
+	re, _ := re(q)
+	r := highlight(q, re, s)
+	if r != h {
+		t.Logf("The highlighting is wrong in ｢%s｣", r)
 		t.Fail()
 	}
 }

index.md

@@ -1,3 +1,5 @@
+# Welcome to Oddµ
+
 Hello! 🙃
 
 Check out the [README](README).

languages.go

@@ -0,0 +1,58 @@
+package main
+
+import (
+	"errors"
+	"github.com/pemistahl/lingua-go"
+	"os"
+	"strings"
+)
+
+// getLangauges returns the environment variable ODDMU_LANGUAGES or
+// all languages.
+func getLanguages() ([]lingua.Language, error) {
+	v := os.Getenv("ODDMU_LANGUAGES")
+	if v == "" {
+		return lingua.AllLanguages(), nil
+	}
+	codes := strings.Split(v, ",")
+	if len(codes) == 1 {
+		return nil, errors.New("detection unnecessary")
+	}
+
+	var langs []lingua.Language
+	for _, lang := range codes {
+		langs = append(langs, lingua.GetLanguageFromIsoCode639_1(lingua.GetIsoCode639_1FromValue(lang)))
+	}
+	return langs, nil
+}
+
+// detector is the LanguageDetector initialized at startup by loadLanguages.
+var detector lingua.LanguageDetector
+
+// loadLanguages initializes the detector using the languages returned
+// by getLanguages and returns the number of languages loaded.
+func loadLanguages() int {
+	langs, err := getLanguages()
+	if err == nil {
+		detector = lingua.NewLanguageDetectorBuilder().
+			FromLanguages(langs...).
+			WithPreloadedLanguageModels().
+			WithLowAccuracyMode().
+			Build()
+	} else {
+		detector = nil
+	}
+	return len(langs)
+}
+
+// language returns the language used for a string, as a lower case
+// ISO 639-1 string, e.g. "en" or "de".
+func language(s string) string {
+	if detector == nil {
+		return os.Getenv("ODDMU_LANGUAGES")
+	}
+	if language, ok := detector.DetectLanguageOf(s); ok {
+		return strings.ToLower(language.IsoCode639_1().String())
+	}
+	return ""
+}

languages_test.go

@@ -0,0 +1,50 @@
+package main
+
+import (
+	"github.com/stretchr/testify/assert"
+	"os"
+	"testing"
+)
+
+func TestAllLanguage(t *testing.T) {
+	os.Unsetenv("ODDMU_LANGUAGES")
+	loadLanguages()
+	l := language(`
+My back hurts at night
+My shoulders won't budge today
+Winter bones I say`)
+	assert.Equal(t, "en", l)
+}
+
+func TestSomeLanguages(t *testing.T) {
+	os.Setenv("ODDMU_LANGUAGES", "en,de")
+	loadLanguages()
+	l := language(`
+Kühle Morgenluft
+Keine Amsel singt heute
+Mensch im Dämmerlicht
+`)
+	assert.Equal(t, "de", l)
+}
+
+func TestOneLanguages(t *testing.T) {
+	os.Setenv("ODDMU_LANGUAGES", "en")
+	loadLanguages()
+	l := language(`
+Schwer wiegt die Luft hier
+Atme ein, ermahn' ich mich
+Erinnerungen
+`)
+	assert.Equal(t, "en", l)
+}
+
+func TestWrongLanguages(t *testing.T) {
+	os.Setenv("ODDMU_LANGUAGES", "de,fr")
+	loadLanguages()
+	l := language(`
+Something drifts down there
+Head submerged oh god a man
+Drowning as we stare
+`)
+	assert.NotEqual(t, "en", l)
+}

oddmu.service

@@ -17,7 +17,6 @@ Environment="ODDMU_PORT=8080"
 ReadWritePaths=/home/oddmu
 ProtectHostname=yes
 RestrictSUIDSGID=yes
-UMask=0077
 RemoveIPC=yes
 MemoryDenyWriteExecute=yes
 

page.go

@@ -1,14 +1,17 @@
 package main
 
 import (
+	"bytes"
+	"fmt"
 	"github.com/gomarkdown/markdown"
 	"github.com/gomarkdown/markdown/ast"
 	"github.com/gomarkdown/markdown/parser"
 	"github.com/microcosm-cc/bluemonday"
 	"html/template"
-	"strings"
-	"bytes"
+	"net/url"
 	"os"
+	"path/filepath"
+	"strings"
 )
 
 // Page is a struct containing information about a single page. Title
@@ -18,11 +21,32 @@ import (
 // page and Html is the rendered HTML for that Markdown. Score is a
 // number indicating how well the page matched for a search query.
 type Page struct {
-	Title string
-	Name  string
-	Body  []byte
-	Html  template.HTML
-	Score int
+	Title    string
+	Name     string
+	Language string
+	Body     []byte
+	Html     template.HTML
+	Score    int
+}
+
+// santize uses bluemonday to sanitize the HTML.
+func sanitize(s string) template.HTML {
+	return template.HTML(bluemonday.UGCPolicy().Sanitize(s))
+}
+
+// santizeBytes uses bluemonday to sanitize the HTML.
+func sanitizeBytes(bytes []byte) template.HTML {
+	return template.HTML(bluemonday.UGCPolicy().SanitizeBytes(bytes))
+}
+
+// nameEscape returns the page name safe for use in URLs. That is,
+// percent escaping is used except for the slashes.
+func nameEscape(s string) string {
+	parts := strings.Split(s, "/")
+	for i, part := range parts {
+		parts[i] = url.PathEscape(part)
+	}
+	return strings.Join(parts, "/")
 }
 
 // save saves a Page. The filename is based on the Page.Name and gets
@@ -33,9 +57,20 @@ type Page struct {
 func (p *Page) save() error {
 	filename := p.Name + ".md"
 	s := bytes.ReplaceAll(p.Body, []byte{'\r'}, []byte{})
+	if len(s) == 0 {
+		return os.Remove(filename)
+	}
 	p.Body = s
 	p.updateIndex()
-	return os.WriteFile(filename, s, 0600)
+	d := filepath.Dir(filename)
+	if d != "." {
+		err := os.MkdirAll(d, 0755)
+		if err != nil {
+			fmt.Printf("Creating directory %s failed", d)
+			return err
+		}
+	}
+	return os.WriteFile(filename, s, 0644)
 }
 
 // loadPage loads a Page given a name. The filename loaded is that
@@ -49,13 +84,13 @@ func loadPage(name string) (*Page, error) {
 	if err != nil {
 		return nil, err
 	}
-	return &Page{Title: name, Name: name, Body: body}, nil
+	return &Page{Title: name, Name: name, Body: body, Language: ""}, nil
 }
 
 // handleTitle extracts the title from a Page and sets Page.Title, if
 // any. If replace is true, the page title is also removed from
 // Page.Body. Make sure not to save this! This is only for rendering.
-func (p* Page) handleTitle(replace bool) {
+func (p *Page) handleTitle(replace bool) {
 	s := string(p.Body)
 	m := titleRegexp.FindStringSubmatch(s)
 	if m != nil {
@@ -66,17 +101,39 @@ func (p* Page) handleTitle(replace bool) {
 	}
 }
 
+func hashtag(p *parser.Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+	i := 0
+	n := len(data)
+	for i < n && !parser.IsSpace(data[i]) {
+		i++
+	}
+	if i == 0 {
+		return 0, nil
+	}
+	link := &ast.Link{
+		Destination: append([]byte("/search?q=%23"), data[1:i]...),
+		Title:       data[0:i],
+	}
+	text := bytes.ReplaceAll(data[0:i], []byte("_"), []byte(" "))
+	ast.AppendChild(link, &ast.Text{Leaf: ast.Leaf{Literal: text}})
+	return i, link
+}
+
 // renderHtml renders the Page.Body to HTML and sets Page.Html.
-func (p* Page) renderHtml() {
-	maybeUnsafeHTML := markdown.ToHTML(p.Body, nil, nil)
-	html := bluemonday.UGCPolicy().SanitizeBytes(maybeUnsafeHTML)
-	p.Html = template.HTML(html);
+func (p *Page) renderHtml() {
+	parser := parser.New()
+	parser.RegisterInline('#', hashtag)
+	maybeUnsafeHTML := markdown.ToHTML(p.Body, parser, nil)
+	p.Name = nameEscape(p.Name)
+	p.Html = sanitizeBytes(maybeUnsafeHTML)
+	p.Language = language(p.plainText())
 }
 
 // plainText renders the Page.Body to plain text and returns it,
 // ignoring all the Markdown and all the newlines. The result is one
 // long single line of text.
-func (p* Page) plainText() string {
+func (p *Page) plainText() string {
 	parser := parser.New()
 	doc := markdown.Parse(p.Body, parser)
 	text := []byte("")
@@ -94,18 +151,17 @@ func (p* Page) plainText() string {
 		}
 	}
 	// Remove trailing space
-	for text[len(text)-1] == ' ' {
-		text = text[0:len(text)-1]
+	for len(text) > 0 && text[len(text)-1] == ' ' {
+		text = text[0 : len(text)-1]
 	}
 	return string(text)
 }
 
 // summarize for query string q sets Page.Html to an extract.
-func (p* Page) summarize(q string) {
+func (p *Page) summarize(q string) {
 	p.handleTitle(true)
-	s, c := snippets(q, p.plainText())
-	p.Score = c
-	extract := []byte(s)
-	html := bluemonday.UGCPolicy().SanitizeBytes(extract)
-	p.Html = template.HTML(html)
+	p.Score = score(q, string(p.Body)) + score(q, p.Title)
+	t := p.plainText()
+	p.Html = sanitize(snippets(q, t))
+	p.Language = language(t)
 }

page_test.go

@@ -1,59 +1,85 @@
 package main
 
 import (
-	"strings"
+	"github.com/stretchr/testify/assert"
+	"os"
+	"regexp"
 	"testing"
 )
 
-func TestPageTitle (t *testing.T) {
+func TestPageTitle(t *testing.T) {
 	p := &Page{Body: []byte(`# Ache
 My back aches for you
 I sit, stare and type for hours
 But yearn for blue sky`)}
 	p.handleTitle(false)
-	if p.Title != "Ache" {
-		t.Logf("The page title was not extracted correctly: %s", p.Title)
-		t.Fail()
-	}
-	if !strings.HasPrefix(string(p.Body), "# Ache") {
-		t.Logf("The page title was removed: %s", p.Body)
-		t.Fail()
-	}
+	assert.Equal(t, "Ache", p.Title)
+	assert.Regexp(t, regexp.MustCompile("^# Ache"), string(p.Body))
 	p.handleTitle(true)
-	if !strings.HasPrefix(string(p.Body), "My back") {
-		t.Logf("The page title was not removed: %s", p.Body)
-		t.Fail()
-	}
+	assert.Regexp(t, regexp.MustCompile("^My back"), string(p.Body))
 }
 
-func TestPagePlainText (t *testing.T) {
+func TestPagePlainText(t *testing.T) {
 	p := &Page{Body: []byte(`# Water
 The air will not come
 To inhale is an effort
 The summer heat kills`)}
-	s := p.plainText()
 	r := "Water The air will not come To inhale is an effort The summer heat kills"
-	if s != r {
-		t.Logf("The plain text version is wrong: %s", s)
-		t.Fail()
-	}
+	assert.Equal(t, r, p.plainText())
 }
 
-func TestPageHtml (t *testing.T) {
+func TestPageHtml(t *testing.T) {
 	p := &Page{Body: []byte(`# Sun
 Silver leaves shine bright
 They droop, boneless, weak and sad
 A cruel sun stares down`)}
 	p.renderHtml()
-	s := string(p.Html)
 	r := `<h1>Sun</h1>
 
 <p>Silver leaves shine bright
 They droop, boneless, weak and sad
 A cruel sun stares down</p>
 `
-	if s != r {
-		t.Logf("The HTML is wrong: %s", s)
-		t.Fail()
-	}
+	assert.Equal(t, r, string(p.Html))
+}
+
+func TestPageHtmlHashtag(t *testing.T) {
+	p := &Page{Body: []byte(`# Comet
+Stars flicker above
+Too faint to focus, so far
+I am cold, alone
+
+#Haiku #Cold_Poets`)}
+	p.renderHtml()
+	r := `<h1>Comet</h1>
+
+<p>Stars flicker above
+Too faint to focus, so far
+I am cold, alone</p>
+
+<p><a href="/search?q=%23Haiku" rel="nofollow">#Haiku</a> <a href="/search?q=%23Cold_Poets" rel="nofollow">#Cold Poets</a></p>
+`
+	assert.Equal(t, r, string(p.Html))
+}
+
+func TestPageDir(t *testing.T) {
+	_ = os.RemoveAll("testdata")
+	loadIndex()
+	p := &Page{Name: "testdata/moon", Body: []byte(`# Moon
+From bed to bathroom
+A slow shuffle in the dark
+Moonlight floods the aisle`)}
+	p.save()
+	o, err := loadPage("testdata/moon")
+	assert.NoError(t, err, "load page")
+	assert.Equal(t, p.Body, o.Body)
+	assert.FileExists(t, "testdata/moon.md")
+
+	// Saving an empty page deletes it.
+	p = &Page{Name: "testdata/moon", Body: []byte("")}
+	p.save()
+	assert.NoFileExists(t, "testdata/moon.md")
+	t.Cleanup(func() {
+		_ = os.RemoveAll("testdata")
+	})
 }

score.go

@@ -0,0 +1,44 @@
+package main
+
+import (
+	"regexp"
+	"strings"
+)
+
+// score splits the query string q into terms and scores the text
+// based on those terms. This assumes that q already has all its meta
+// characters quoted.
+func score(q string, s string) int {
+	score := 0
+	re, err := regexp.Compile("(?i)" + q)
+	if err == nil {
+		m := re.FindAllString(s, -1)
+		if m != nil {
+			// Score increases for each full match of q.
+			score += len(m)
+		}
+	}
+	for _, v := range strings.Fields(q) {
+		re, err := regexp.Compile(`(?is)(\pL?)(` + v + `)(\pL?)`)
+		if err != nil {
+			continue
+		}
+		for _, m := range re.FindAllStringSubmatch(s, -1) {
+			// Term matched increases the score.
+			score++
+			// Terms matching at the beginning and
+			// end of words and matching entire
+			// words increase the score further.
+			if len(m[1]) == 0 {
+				score++
+			}
+			if len(m[3]) == 0 {
+				score++
+			}
+			if len(m[1]) == 0 && len(m[3]) == 0 {
+				score++
+			}
+		}
+	}
+	return score
+}

score_test.go

@@ -0,0 +1,104 @@
+package main
+
+import (
+	"testing"
+)
+
+func TestScore(t *testing.T) {
+
+	s := `The windows opens
+A wave of car noise hits me
+No birds to be heard.`
+
+	q := "window"
+	// Score:
+	// - q itself
+	// - the single token
+	// - the beginning of a word
+	c := score(q, s)
+	if c != 3 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+	q = "windows"
+	c = score(q, s)
+	// Score:
+	// - q itself
+	// - the single token
+	// - the beginning of a word
+	// - the end of a word
+	// - the whole word
+	if c != 5 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+	q = "car noise"
+	c = score(q, s)
+	// Score:
+	// - car noise (+1)
+	// - car, with beginning, end, whole word (+4)
+	// - noise, with beginning, end, whole word (+4)
+	if c != 9 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+	q = "noise car"
+	c = score(q, s)
+	// Score:
+	// - the car token
+	// - the noise token
+	// - each with beginning, end and whole token (3 each)
+	if c != 8 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+}
+
+func TestScoreLong(t *testing.T) {
+	s := `We are immersed in a sea of dead people. All the dead that have gone before us, silent now, just staring, gaping. As we move and talk and fret, never once stopping to ask ourselves – or them! – what it was all about. Instead we drown ourselves in noise. Incessantly we babble, surrounded by false friends claiming that all is well. And look at us! Yes, we are well. Patting our backs and expecting a pat – and we do! – we smugly do enjoy.`
+	q := "all is well"
+	c := score(q, s)
+	// Score:
+	// - all is well (1)
+	// - all, beginning, end, whole word (+4 × 3 = 12)
+	// - is, beginning, end, whole word (+4 × 1 = 4), and as a substring (1)
+	// - well, beginning, end, whole word (+4 × 2 = 8)
+	if c != 26 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+}
+
+func TestScoreSubstring(t *testing.T) {
+	s := `The loneliness of space means that receiving messages means knowledge that other people are out there. Not satellites pinging forever. Not bots searching and probing. Instead, humans. People who care. Curious and cautious.`
+	q := "search probe"
+	c := score(q, s)
+	// Score:
+	// - search, beginning (2)
+	// - probe (0)
+	if c != 2 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+	q = "ear"
+	c = score(q, s)
+	// Score:
+	// - ear, all (2)
+	if c != 2 {
+		t.Logf("%s score is %d", q, c)
+		t.Fail()
+	}
+}
+
+func TestScorePageAndMarkup(t *testing.T) {
+	s := `The Transjovian Council accepts new members. If you think we'd be a good fit, apply for an account. Contact [Alex Schroeder](https://alexschroeder.ch/wiki/Contact). Mail is best. Encrypted mail is best. [Delta Chat](https://delta.chat/de/) is a messenger app that uses encrypted mail. It's the bestest best.`
+	p := &Page{Title: "Test", Name: "Test", Body: []byte(s)}
+	q := "wiki"
+	p.summarize(q)
+	// "wiki" is not visible in the plain text but the score is no affected:
+	// - wiki, all, whole, beginning, end (5)
+	if p.Score != 5 {
+		t.Logf("%s score is %d", q, p.Score)
+		t.Fail()
+	}
+}

search.go

@@ -1,12 +1,15 @@
 package main
 
 import (
-	trigram "github.com/dgryski/go-trigram"
-	"path/filepath"
-	"strings"
-	"slices"
-	"io/fs"
 	"fmt"
+	trigram "github.com/dgryski/go-trigram"
+	"io/fs"
+	"path/filepath"
+	"slices"
+	"strings"
+	"sync"
+	"unicode"
+	"unicode/utf8"
 )
 
 // Search is a struct containing the result of a search. Query is the
@@ -19,14 +22,23 @@ type Search struct {
 	Results bool
 }
 
-// index is a struct containing the trigram index for search. It is
-// generated at startup and updated after every page edit.
-var index trigram.Index
+// idx contains the two maps used for search. Make sure to lock and
+// unlock as appropriate.
+var idx = struct {
+	sync.RWMutex
 
-// documents is a map, mapping document ids of the index to page
-// names.
-var documents map[trigram.DocID]string
+	// index is a struct containing the trigram index for search. It is
+	// generated at startup and updated after every page edit. The index
+	// is case-insensitive.
+	index trigram.Index
 
+	// documents is a map, mapping document ids of the index to page
+	// names.
+	documents map[trigram.DocID]string
+}{}
+
+// indexAdd reads a file and adds it to the index. This must happen
+// while the idx is locked, which is true when called from loadIndex.
 func indexAdd(path string, info fs.FileInfo, err error) error {
 	if err != nil {
 		return err
@@ -40,50 +52,91 @@ func indexAdd(path string, info fs.FileInfo, err error) error {
 	if err != nil {
 		return err
 	}
-	id := index.Add(string(p.Body))
-	documents[id] = p.Name
+	id := idx.index.Add(strings.ToLower(string(p.Body)))
+	idx.documents[id] = p.Name
 	return nil
 }
 
-func loadIndex() error {
-	index = make(trigram.Index)
-	documents = make(map[trigram.DocID]string)
+// loadIndex loads all the pages and indexes them. This takes a while.
+// It returns the number of pages indexed.
+func loadIndex() (int, error) {
+	idx.Lock()
+	defer idx.Unlock()
+	idx.index = make(trigram.Index)
+	idx.documents = make(map[trigram.DocID]string)
 	err := filepath.Walk(".", indexAdd)
 	if err != nil {
-		fmt.Println("Indexing failed")
-		index = nil
-		documents = nil
+		idx.index = nil
+		idx.documents = nil
+		return 0, err
 	}
-	return err
+	n := len(idx.documents)
+	return n, nil
 }
 
+// updateIndex updates the index for a single page. The old text is
+// loaded from the disk and removed from the index first, if it
+// exists.
 func (p *Page) updateIndex() {
+	idx.Lock()
+	defer idx.Unlock()
 	var id trigram.DocID
-	for docId, name := range documents {
+	// This function does not rely on files actually existing, so
+	// let's quickly find the document id.
+	for docId, name := range idx.documents {
 		if name == p.Name {
 			id = docId
 			break
 		}
 	}
 	if id == 0 {
-		id = index.Add(string(p.Body))
-		documents[id] = p.Name
+		id = idx.index.Add(strings.ToLower(string(p.Body)))
+		idx.documents[id] = p.Name
 	} else {
 		o, err := loadPage(p.Name)
 		if err == nil {
-			index.Delete(string(o.Body), id)
+			idx.index.Delete(strings.ToLower(string(o.Body)), id)
 		}
-		index.Insert(string(p.Body), id)
+		idx.index.Insert(strings.ToLower(string(p.Body)), id)
 	}
 }
 
-// search returns a sorted []Page where each page contains an extract
-// of the actual Page.Body in its Page.Html.
-func search(q string) []Page {
-	ids := index.Query(q)
-	items := make([]Page, len(ids))
-	for i, id := range ids {
-		name := documents[id]
+func sortItems(a, b Page) int {
+	// Sort by score
+	if a.Score < b.Score {
+		return 1
+	} else if a.Score > b.Score {
+		return -1
+	}
+	// If the score is the same and both page names start
+	// with a number (like an ISO date), sort descending.
+	ra, _ := utf8.DecodeRuneInString(a.Title)
+	rb, _ := utf8.DecodeRuneInString(b.Title)
+	if unicode.IsNumber(ra) && unicode.IsNumber(rb) {
+		if a.Title < b.Title {
+			return 1
+		} else if a.Title > b.Title {
+			return -1
+		} else {
+			return 0
+		}
+	}
+	// Otherwise sort ascending.
+	if a.Title < b.Title {
+		return -1
+	} else if a.Title > b.Title {
+		return 1
+	} else {
+		return 0
+	}
+}
+
+// loadAndSummarize loads the pages named and summarizes them for the
+// query give.
+func loadAndSummarize(names []string, q string) []Page {
+	// Load and summarize the items.
+	items := make([]Page, len(names))
+	for i, name := range names {
 		p, err := loadPage(name)
 		if err != nil {
 			fmt.Printf("Error loading %s\n", name)
@@ -92,19 +145,30 @@ func search(q string) []Page {
 			items[i] = *p
 		}
 	}
-	fn := func(a, b Page) int {
-		if a.Score < b.Score {
-			return 1
-		} else if a.Score > b.Score {
-			return -1
-		} else if a.Title < b.Title {
-			return -1
-		} else if a.Title > b.Title {
-			return 1
-		} else {
-			return 0
-		}
-	}
-	slices.SortFunc(items, fn)
+	return items
+}
+
+// search returns a sorted []Page where each page contains an extract
+// of the actual Page.Body in its Page.Html.
+func search(q string) []Page {
+	if len(q) == 0 {
+		return make([]Page, 0)
+	}
+	words := strings.Fields(strings.ToLower(q))
+	var trigrams []trigram.T
+	for _, word := range words {
+		trigrams = trigram.Extract(word, trigrams)
+	}
+	// Keep the read lock for a short as possible. Make a list of
+	// the names we need to load and summarize.
+	idx.RLock()
+	ids := idx.index.QueryTrigrams(trigrams)
+	names := make([]string, len(ids))
+	for i, id := range ids {
+		names[i] = idx.documents[id]
+	}
+	idx.RUnlock()
+	items := loadAndSummarize(names, q)
+	slices.SortFunc(items, sortItems)
 	return items
 }

search.html

@@ -6,23 +6,37 @@
     <meta name="viewport" content="width=device-width">
     <title>Search for {{.Query}}</title>
     <style>
-html { max-width: 70ch; padding: 2ch; margin: auto; color: #111; background: #ffe; }
+html { max-width: 65ch; padding: 2ch; margin: auto; color: #111; background: #ffe; }
+body { hyphens: auto; }
+header a { margin-right: 1ch; }
+form { display: inline-block; }
 img { max-width: 20%; }
 .result { font-size: larger }
 .score { font-size: smaller; opacity: 0.8; }
     </style>
   </head>
-  <body>
-    <h1>Search for {{.Query}}</h1>
-    <div>
+  <body lang="en">
+    <header>
+      <a href="#main">Skip navigation</a>
+      <a href="/view/index">Home</a>
+      <form role="search" action="/search" method="GET">
+        <input type="text" value="{{.Query}}" spellcheck="false" name="q" required>
+        <button>Search</button>
+      </form>
+    </header>
+    <main id="main">
+      <h1>Search for {{.Query}}</h1>
       {{if .Results}}
-        {{range .Items}}
-      <p><a class="result" href="/view/{{.Name}}">{{.Title}}</a> <span class="score">{{.Score}}</span></p>
-      <blockquote>{{.Html}}</blockquote>
-        {{end}}
+      {{range .Items}}
+      <article lang="{{.Language}}">
+        <p><a class="result" href="/view/{{.Name}}">{{.Title}}</a>
+          <span class="score">{{.Score}}</span></p>
+        <blockquote>{{.Html}}</blockquote>
+      </article>
+      {{end}}
       {{else}}
       <p>No results.</p>
       {{end}}
-    </div>
+    </main>
   </body>
 </html>

search_test.go

@@ -1,36 +1,41 @@
 package main
 
 import (
-	"testing"
-	"strings"
+	"github.com/stretchr/testify/assert"
 	"os"
+	"strings"
+	"testing"
 )
 
-var name string = "test"
-
 // TestIndex relies on README.md being indexed
-func TestIndex (t *testing.T) {
-	_ = os.Remove(name + ".md")
+func TestIndex(t *testing.T) {
 	loadIndex()
 	q := "Oddµ"
 	pages := search(q)
-	if len(pages) == 0 {
-		t.Log("Search found no result")
-		t.Fail()
-	}
+	assert.NotZero(t, len(pages))
 	for _, p := range pages {
-		if !strings.Contains(string(p.Body), q) {
-			t.Logf("Page %s does not contain %s", p.Name, q)
-			t.Fail()
-		}
-		if p.Score == 0 {
-			t.Logf("Page %s has no score", p.Name)
-			t.Fail()
-		}
+		assert.NotContains(t, p.Title, "<b>")
+		assert.True(t, strings.Contains(string(p.Body), q) || strings.Contains(string(p.Title), q))
+		assert.NotZero(t, p.Score)
 	}
+}
+
+func TestSearchHashtag(t *testing.T) {
+	loadIndex()
+	q := "#Another_Tag"
+	pages := search(q)
+	assert.NotZero(t, len(pages))
+}
+
+func TestIndexUpdates(t *testing.T) {
+	name := "test"
+	_ = os.Remove(name + ".md")
+	loadIndex()
 	p := &Page{Name: name, Body: []byte("This is a test.")}
 	p.save()
-	pages = search("This is a test")
+
+	// Find the phrase
+	pages := search("This is a test")
 	found := false
 	for _, p := range pages {
 		if p.Name == name {
@@ -38,10 +43,31 @@ func TestIndex (t *testing.T) {
 			break
 		}
 	}
-	if !found {
-		t.Logf("Page '%s' was not found", name)
-		t.Fail()
+	assert.True(t, found)
+
+	// Find the phrase, case insensitive
+	pages = search("this is a test")
+	found = false
+	for _, p := range pages {
+		if p.Name == name {
+			found = true
+			break
+		}
 	}
+	assert.True(t, found)
+
+	// Find some words
+	pages = search("this test")
+	found = false
+	for _, p := range pages {
+		if p.Name == name {
+			found = true
+			break
+		}
+	}
+	assert.True(t, found)
+
+	// Update the page and no longer find it with the old phrase
 	p = &Page{Name: name, Body: []byte("Guvf vf n grfg.")}
 	p.save()
 	pages = search("This is a test")
@@ -52,10 +78,9 @@ func TestIndex (t *testing.T) {
 			break
 		}
 	}
-	if found {
-		t.Logf("Page '%s' was still found using the old content: %s", name, p.Body)
-		t.Fail()
-	}
+	assert.False(t, found)
+
+	// Find page using a new word
 	pages = search("Guvf")
 	found = false
 	for _, p := range pages {
@@ -64,8 +89,9 @@ func TestIndex (t *testing.T) {
 			break
 		}
 	}
-	if !found {
-		t.Logf("Page '%s' not found using the new content: %s", name, p.Body)
-		t.Fail()
-	}
+	assert.True(t, found)
+
+	t.Cleanup(func() {
+		_ = os.Remove(name + ".md")
+	})
 }

snippets.go

@@ -1,22 +1,36 @@
 package main
 
 import (
-	"strings"
 	"regexp"
+	"strings"
 )
 
-func snippets (q string, s string) (string, int) {
+// re returns a regular expression matching any word in q.
+func re(q string) (*regexp.Regexp, error) {
+	q = regexp.QuoteMeta(q)
+	re, err := regexp.Compile(`\s+`)
+	if err != nil {
+		return nil, err
+	}
+	words := re.ReplaceAllString(q, "|")
+	re, err = regexp.Compile(`(?i)(` + words + `)`)
+	if err != nil {
+		return nil, err
+	}
+	return re, nil
+}
+
+func snippets(q string, s string) string {
 	// Look for Snippets
 	snippetlen := 100
 	maxsnippets := 4
-	// Compile the query as a regular expression
-	re, err := regexp.Compile("(?i)(" + strings.Join(strings.Split(q, " "), "|") + ")")
-	// If the compilation didn't work, truncate
+	re, err := re(q)
+	// If the compilation didn't work, truncate and return
 	if err != nil || len(s) <= snippetlen {
 		if len(s) > 400 {
-			s = s[0:400]
+			s = s[0:400] + " …"
 		}
-		return highlight(q, s)
+		return s
 	}
 	// show a snippet from the beginning of the document
 	j := strings.LastIndex(s[:snippetlen], " ")
@@ -26,9 +40,9 @@ func snippets (q string, s string) (string, int) {
 		if j == -1 {
 			// Or just truncate the body.
 			if len(s) > 400 {
-				s = s[0:400]
+				s = s[0:400] + " …"
 			}
-			return highlight(q, s)
+			return highlight(q, re, s)
 		}
 	}
 	t := s[0:j]
@@ -45,7 +59,7 @@ func snippets (q string, s string) (string, int) {
 		if j > -1 {
 			// get the substring containing the start of
 			// the match, ending on word boundaries
-			from := j - snippetlen / 2
+			from := j - snippetlen/2
 			if from < 0 {
 				from = 0
 			}
@@ -55,7 +69,7 @@ func snippets (q string, s string) (string, int) {
 			} else {
 				start += from
 			}
-			to := j + snippetlen / 2
+			to := j + snippetlen/2
 			if to > len(s) {
 				to = len(s)
 			}
@@ -69,11 +83,11 @@ func snippets (q string, s string) (string, int) {
 					end += to
 				}
 			}
-			t = s[start : end];
-			res = res + t + " …";
+			t = s[start:end]
+			res = res + t + " …"
 			// truncate text to avoid rematching the same string.
 			s = s[end:]
 		}
 	}
-	return highlight(q, res)
+	return highlight(q, re, res)
 }

snippets_test.go

@@ -10,18 +10,9 @@ func TestSnippets(t *testing.T) {
 	h := `We are immersed in a sea of dead people. <b>All</b> the dead that have gone before us, silent now, just … to ask ourselves – or them! – what it was <b>all</b> about. Instead we drown ourselves in no<b>is</b>e. … surrounded by false friends claiming that <b>all</b> <b>is</b> <b>well</b>. And look at us! Yes, we are <b>well</b>. …`
 
 	q := "all is well"
-	r, c := snippets(q, s)
+	r := snippets(q, s)
 	if r != h {
 		t.Logf("The snippets are wrong in ｢%s｣", r)
 		t.Fail()
 	}
-	// Score 12:
-	// - all is well (1)
-	// - all, beginning, end, whole word (+4 × 3 = 12)
-	// - is, beginning, end, whole word (+4 × 1 = 4), and as a substring (1)
-	// - well, beginning, end, whole word (+4 × 2 = 8)
-	if c != 26 {
-		t.Logf("%s score is %d", q, c)
-		t.Fail()
-	}
 }

upload.html

@@ -0,0 +1,22 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <meta charset="utf-8">
+    <meta name="format-detection" content="telephone=no">
+    <meta name="viewport" content="width=device-width">
+    <title>Upload File</title>
+    <style>
+html { max-width: 70ch; padding: 2ch; margin: auto; color: #111; background: #ffe; }
+form, textarea { width: 100%; }
+    </style>
+  </head>
+  <body>
+    <h1>Upload File</h1>
+    <form action="/save" method="POST" enctype="multipart/form-data">
+      <input type="text" name="name" placeholder="image.jpg" autofocus required>
+      <p><input type="file" name="file" required>
+      <p><input type="submit" value="Save">
+        <a href="/view/index"><button type="button">Cancel</button></a></p>
+    </form>
+  </body>
+</html>

view.html

@@ -6,22 +6,33 @@
     <meta name="viewport" content="width=device-width">
     <title>{{.Title}}</title>
     <style>
-html { max-width: 70ch; padding: 2ch; margin: auto; color: #111; background: #ffe; }
-form { display: inline-block; padding-left: 1em; }
+html { max-width: 65ch; padding: 1ch; margin: auto; color: #111; background: #ffe; }
+body { hyphens: auto; }
+header a { margin-right: 1ch; }
+form { display: inline-block; }
+footer { border-top: 1px solid #888 }
 img { max-width: 100%; }
     </style>
   </head>
-  <body>
-    <h1>{{.Title}}</h1>
-    <div>
-      <a href="/edit/{{.Name}}">Edit this page</a>
+  <body lang="{{.Language}}">
+    <header>
+      <a href="#main">Skip navigation</a>
+      <a href="/view/index">Home</a>
+      <a href="/edit/{{.Name}}">Edit</a>
+      <a href="/add/{{.Name}}">Add</a>
       <form role="search" action="/search" method="GET">
         <input type="text" spellcheck="false" name="q" required>
         <button>Search</button>
       </form>
-    </div>
-    <div>
-{{.Html}}
-    </div>
+    </header>
+    <main id="main">
+      <h1>{{.Title}}</h1>
+      {{.Html}}
+    </main>
+    <footer>
+      <address>
+        Comments? Send mail to Alex Schroeder <<a href="mailto:alex@alexschroeder.ch">alex@alexschroeder.ch</a>>
+      </address>
+    </footer>
   </body>
 </html>

wiki.go

@@ -1,22 +1,25 @@
 package main
 
 import (
-	"html/template"
-	"net/http"
-	"strings"
-	"regexp"
 	"fmt"
+	"html/template"
+	"io"
+	"net/http"
 	"os"
+	"path/filepath"
+	"regexp"
 )
 
 // Templates are parsed at startup.
-var templates = template.Must(template.ParseFiles("edit.html", "view.html", "search.html"))
+var templates = template.Must(
+	template.ParseFiles("edit.html", "add.html", "view.html",
+		"search.html", "upload.html"))
 
 // validPath is a regular expression where the second group matches a
-// page, so when the handler for "/edit/" is called, a URL path of
-// "/edit/foo" results in the editHandler being called with title
-// "foo". The regular expression doesn't define the handlers (this
-// happens in the main function).
+// page, so when the editHandler is called, a URL path of "/edit/foo"
+// results in the editHandler being called with title "foo". The
+// regular expression doesn't define the handlers (this happens in the
+// main function).
 var validPath = regexp.MustCompile("^/([^/]+)/(.+)$")
 
 // titleRegexp is a regular expression matching a level 1 header line
@@ -39,33 +42,31 @@ func rootHandler(w http.ResponseWriter, r *http.Request) {
 	http.Redirect(w, r, "/view/index", http.StatusFound)
 }
 
-// viewHandler renders a text file, if the name ends in ".txt" and
-// such a file exists. Otherwise, it loads the page. If this didn't
-// work, the browser is redirected to an edit page. Otherwise, the
-// "view.html" template is used to show the rendered HTML.
+// viewHandler serves existing files (including markdown files with
+// the .md extension). If the requested file does not exist, a page
+// with the same name is loaded. This means adding the .md extension
+// and using the "view.html" template to render the HTML. Both
+// attempts fail, the browser is redirected to an edit page.
 func viewHandler(w http.ResponseWriter, r *http.Request, name string) {
-	// Short cut for text files
-	if (strings.HasSuffix(name, ".txt")) {
-		body, err := os.ReadFile(name)
-		if err == nil {
-			w.Write(body)
-			return
-		}
-	}
-	// Attempt to load Markdown page; edit it if this fails
-	p, err := loadPage(name)
-	if err != nil {
-		http.Redirect(w, r, "/edit/"+name, http.StatusFound)
+	body, err := os.ReadFile(name)
+	if err == nil {
+		w.Write(body)
 		return
 	}
-	p.handleTitle(true)
-	p.renderHtml()
-	renderTemplate(w, "view", p)
+	p, err := loadPage(name)
+	if err == nil {
+		p.handleTitle(true)
+		p.renderHtml()
+		renderTemplate(w, "view", p)
+		return
+	}
+	http.Redirect(w, r, "/edit/"+name, http.StatusFound)
 }
 
 // editHandler uses the "edit.html" template to present an edit page.
 // When editing, the page title is not overriden by a title in the
-// text. Instead, the page name is used.
+// text. Instead, the page name is used. The edit is saved using the
+// saveHandler.
 func editHandler(w http.ResponseWriter, r *http.Request, name string) {
 	p, err := loadPage(name)
 	if err != nil {
@@ -89,12 +90,89 @@ func saveHandler(w http.ResponseWriter, r *http.Request, name string) {
 	http.Redirect(w, r, "/view/"+name, http.StatusFound)
 }
 
+// addHandler uses the "add.html" template to present an empty edit
+// page. What you type there is appended to the page using the
+// appendHandler.
+func addHandler(w http.ResponseWriter, r *http.Request, name string) {
+	p, err := loadPage(name)
+	if err != nil {
+		p = &Page{Title: name, Name: name}
+	} else {
+		p.handleTitle(false)
+	}
+	renderTemplate(w, "add", p)
+}
+
+// appendHandler takes the "body" form parameter and appends it. The
+// browser is redirected to the page view.
+func appendHandler(w http.ResponseWriter, r *http.Request, name string) {
+	body := r.FormValue("body")
+	p, err := loadPage(name)
+	if err != nil {
+		p = &Page{Title: name, Name: name, Body: []byte(body)}
+	} else {
+		p.handleTitle(false)
+		p.Body = append(p.Body, []byte(body)...)
+	}
+	err = p.save()
+	if err != nil {
+		http.Error(w, err.Error(), http.StatusInternalServerError)
+		return
+	}
+	http.Redirect(w, r, "/view/"+name, http.StatusFound)
+}
+
+// uploadHandler uses the "upload.html" template to enable uploads.
+// The file is saved using the saveUploadHandler.
+func uploadHandler(w http.ResponseWriter, r *http.Request) {
+	renderTemplate(w, "upload", nil)
+}
+
+// saveUploadHandler takes the "name" form field and the "file" form
+// file and saves the file under the given name. The browser is
+// redirected to the view of that file.
+func saveUploadHandler(w http.ResponseWriter, r *http.Request) {
+	filename := r.FormValue("name")
+	file, _, err := r.FormFile("file")
+	if err != nil {
+		http.Error(w, err.Error(), http.StatusInternalServerError)
+		return
+	}
+	defer file.Close()
+	// backup an existing file with the same name
+	_, err = os.Stat(filename)
+	if err != nil {
+		os.Rename(filename, filename + "~")
+	}
+	// create the directory, if necessary
+	d := filepath.Dir(filename)
+	if d != "." {
+		err := os.MkdirAll(d, 0755)
+		if err != nil {
+			http.Error(w, err.Error(), http.StatusInternalServerError)
+			return
+		}
+	}
+	// create the new file
+	dst, err := os.Create(filename)
+	if err != nil {
+            http.Error(w, err.Error(), http.StatusInternalServerError)
+            return
+        }
+	defer dst.Close()
+	if _, err := io.Copy(dst, file); err != nil {
+            http.Error(w, err.Error(), http.StatusInternalServerError)
+            return
+        }
+	http.Redirect(w, r, "/view/"+filename, http.StatusFound)
+}
+
 // makeHandler returns a handler that uses the URL path without the
 // first path element as its argument, e.g. if the URL path is
 // /edit/foo/bar, the editHandler is called with "foo/bar" as its
 // argument. This uses the second group from the validPath regular
 // expression.
-func makeHandler(fn func (http.ResponseWriter, *http.Request, string)) http.HandlerFunc {
+func makeHandler(fn func(http.ResponseWriter, *http.Request, string)) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		m := validPath.FindStringSubmatch(r.URL.Path)
 		if m != nil {
@@ -125,14 +203,49 @@ func getPort() string {
 	return port
 }
 
-func main() {
+// scheduleLoadIndex calls loadIndex and prints some messages before
+// and after. For testing, call loadIndex directly and skip the
+// messages.
+func scheduleLoadIndex() {
+	fmt.Print("Indexing pages\n")
+	n, err := loadIndex()
+	if err == nil {
+		fmt.Printf("Indexed %d pages\n", n)
+	} else {
+		fmt.Println("Indexing failed")
+	}
+}
+
+// scheduleLoadLanguages calls loadLanguages and prints some messages before
+// and after. For testing, call loadLanguages directly and skip the
+// messages.
+func scheduleLoadLanguages() {
+	fmt.Print("Loading languages\n")
+	n := loadLanguages()
+	fmt.Printf("Loaded %d languages\n", n)
+}
+
+func serve() {
 	http.HandleFunc("/", rootHandler)
 	http.HandleFunc("/view/", makeHandler(viewHandler))
 	http.HandleFunc("/edit/", makeHandler(editHandler))
 	http.HandleFunc("/save/", makeHandler(saveHandler))
+	http.HandleFunc("/add/", makeHandler(addHandler))
+	http.HandleFunc("/append/", makeHandler(appendHandler))
+	http.HandleFunc("/upload", uploadHandler)
+	http.HandleFunc("/save", saveUploadHandler)
 	http.HandleFunc("/search", searchHandler)
-	loadIndex()
+	go scheduleLoadIndex()
+	go scheduleLoadLanguages()
 	port := getPort()
 	fmt.Printf("Serving a wiki on port %s\n", port)
-	http.ListenAndServe(":" + port, nil)
+	http.ListenAndServe(":"+port, nil)
+}
+
+func main() {
+	if len(os.Args) == 1 {
+		serve()
+	} else {
+		commands()
+	}
 }

wiki_test.go

@@ -0,0 +1,184 @@
+package main
+
+import (
+	"bytes"
+	"github.com/stretchr/testify/assert"
+	"mime/multipart"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"os"
+	"regexp"
+	"strings"
+	"testing"
+)
+
+// HTTPHeaders is a helper that returns HTTP headers of the response. It returns
+// nil if building a new request fails.
+func HTTPHeaders(handler http.HandlerFunc, method, url string, values url.Values, header string) []string {
+	w := httptest.NewRecorder()
+	req, err := http.NewRequest(method, url+"?"+values.Encode(), nil)
+	if err != nil {
+		return nil
+	}
+	handler(w, req)
+	return w.Result().Header[header]
+}
+
+// HTTPRedirectTo checks that the request results in a redirect and it
+// checks the destination of the redirect. It returns whether the
+// request did in fact result in a redirect. Note: This method assumes
+// that POST requests ignore the query part of the URL.
+func HTTPRedirectTo(t *testing.T, handler http.HandlerFunc, method, url string, values url.Values, destination string) bool {
+	w := httptest.NewRecorder()
+	var req *http.Request
+	var err error
+	if method == http.MethodPost {
+		body := strings.NewReader(values.Encode())
+		req, err = http.NewRequest(method, url, body)
+		req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
+	} else {
+		req, err = http.NewRequest(method, url+"?"+values.Encode(), nil)
+	}
+	assert.NoError(t, err)
+	handler(w, req)
+	code := w.Code
+	isRedirectCode := code >= http.StatusMultipleChoices && code <= http.StatusTemporaryRedirect
+	assert.True(t, isRedirectCode, "Expected HTTP redirect status code for %q but received %d", url+"?"+values.Encode(), code)
+	headers := w.Result().Header["Location"]
+	assert.True(t, len(headers) == 1 && headers[0] == destination,
+		"Expected HTTP redirect location %s for %q but received %v", destination, url+"?"+values.Encode(), headers)
+	return isRedirectCode
+}
+
+// HTTPUploadAndRedirectTo checks that the request results in a redirect and it
+// checks the destination of the redirect. It returns whether the
+// request did in fact result in a redirect.
+func HTTPUploadAndRedirectTo(t *testing.T, handler http.HandlerFunc, url, contentType string, body *bytes.Buffer, destination string) bool {
+	w := httptest.NewRecorder()
+	req, err := http.NewRequest("POST", url, body)
+	req.Header.Set("Content-Type", contentType)
+	assert.NoError(t, err)
+	handler(w, req)
+	code := w.Code
+	isRedirectCode := code >= http.StatusMultipleChoices && code <= http.StatusTemporaryRedirect
+	assert.True(t, isRedirectCode, "Expected HTTP redirect status code for %q but received %d", url, code)
+	headers := w.Result().Header["Location"]
+	assert.True(t, len(headers) == 1 && headers[0] == destination,
+		"Expected HTTP redirect location %s for %q but received %v", destination, url, headers)
+	return isRedirectCode
+}
+
+func TestRootHandler(t *testing.T) {
+	HTTPRedirectTo(t, rootHandler, "GET", "/", nil, "/view/index")
+}
+
+// relies on index.md in the current directory!
+func TestViewHandler(t *testing.T) {
+	assert.Regexp(t, regexp.MustCompile("Welcome to Oddµ"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/index", nil))
+}
+
+// wipes testdata
+func TestEditSave(t *testing.T) {
+	_ = os.RemoveAll("testdata")
+
+	data := url.Values{}
+	data.Set("body", "Hallo!")
+
+	HTTPRedirectTo(t, makeHandler(viewHandler), "GET", "/view/testdata/alex", nil, "/edit/testdata/alex")
+	assert.HTTPStatusCode(t, makeHandler(editHandler), "GET", "/edit/testdata/alex", nil, 200)
+	HTTPRedirectTo(t, makeHandler(saveHandler), "POST", "/save/testdata/alex", data, "/view/testdata/alex")
+	assert.Regexp(t, regexp.MustCompile("Hallo!"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/alex", nil))
+
+	t.Cleanup(func() {
+		_ = os.RemoveAll("testdata")
+	})
+}
+
+// wipes testdata
+func TestAddAppend(t *testing.T) {
+	_ = os.RemoveAll("testdata")
+
+	p := &Page{Name: "testdata/fire", Body: []byte(`# Fire
+Orange sky above
+Reflects a distant fire
+It's not `)}
+	p.save()
+
+	data := url.Values{}
+	data.Set("body", "barbecue")
+
+	assert.Regexp(t, regexp.MustCompile("a distant fire"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/fire", nil))
+	assert.NotRegexp(t, regexp.MustCompile("a distant fire"),
+		assert.HTTPBody(makeHandler(addHandler), "GET", "/add/testdata/fire", nil))
+	HTTPRedirectTo(t, makeHandler(appendHandler), "POST", "/append/testdata/fire", data, "/view/testdata/fire")
+	assert.Regexp(t, regexp.MustCompile("It’s not barbecue"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/fire", nil))
+
+	t.Cleanup(func() {
+		_ = os.RemoveAll("testdata")
+	})
+}
+
+// wipes testdata
+func TestUpload(t *testing.T) {
+	_ = os.RemoveAll("testdata")
+	assert.HTTPStatusCode(t, uploadHandler, "GET", "/upload", nil, 200)
+	form := new(bytes.Buffer)
+	writer := multipart.NewWriter(form)
+	field, err := writer.CreateFormField("name")
+	assert.NoError(t, err)
+	_, err = field.Write([]byte("testdata/ok.txt"))
+	assert.NoError(t, err)
+	file, err := writer.CreateFormFile("file", "example.txt");
+	assert.NoError(t, err)
+	file.Write([]byte("Hello!"))
+	err = writer.Close()
+	assert.NoError(t, err)
+	t.Log(writer.FormDataContentType())
+	HTTPUploadAndRedirectTo(t, saveUploadHandler, "/upload", writer.FormDataContentType(), form, "/view/testdata/ok.txt")
+	assert.Regexp(t, regexp.MustCompile("Hello!"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/ok.txt", nil))
+	t.Cleanup(func() {
+		_ = os.RemoveAll("testdata")
+	})
+}
+
+// wipes testdata
+func TestPageTitleWithAmp(t *testing.T) {
+	_ = os.RemoveAll("testdata")
+
+	p := &Page{Name: "testdata/Rock & Roll", Body: []byte("Dancing")}
+	p.save()
+
+	assert.Regexp(t, regexp.MustCompile("Rock &amp; Roll"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/Rock%20%26%20Roll", nil))
+
+	p = &Page{Name: "testdata/Rock & Roll", Body: []byte("# Sex & Drugs & Rock'n'Roll\nOh no!")}
+	p.save()
+
+	assert.Regexp(t, regexp.MustCompile("Sex &amp; Drugs"),
+		assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/Rock%20%26%20Roll", nil))
+
+	t.Cleanup(func() {
+		_ = os.RemoveAll("testdata")
+	})
+}
+
+func TestPageTitleWithQuestionMark(t *testing.T) {
+	_ = os.RemoveAll("testdata")
+
+	p := &Page{Name: "testdata/How about no?", Body: []byte("No means no")}
+	p.save()
+
+	body := assert.HTTPBody(makeHandler(viewHandler), "GET", "/view/testdata/How%20about%20no%3F", nil)
+	assert.Contains(t, body, "No means no")
+	assert.Contains(t, body, "<a href=\"/edit/testdata/How%20about%20no%3F\">Edit</a>")
+
+	t.Cleanup(func() {
+		_ = os.RemoveAll("testdata")
+	})
+}