Documentation
¶
Overview ¶
Command deployid echoes a deploy identifier from the environment into a meta tag, and says when it could not put it somewhere a browser will read.
$ DEPLOY_ID=d7f3a91 deployid page.html <!doctype html><meta name="deploy-id" content="d7f3a91"><html><head>… $ DEPLOY_ID=d7f3a91 deployid -report page.html the meta is in the head anchor before the first element, <html> head reached yes
A comment can go anywhere. A meta has to be in the head or a browser ignores it, and a rewriter cannot see the head that a parser will build - see examples/gip/buildinfo and B182. So the question is not where to put the tag but whether the place available is a place that counts.
The parser builds the head around the insertion, unless text got there first ¶
Inserting a bare meta before the first element is enough: a parser in its "before head" mode meets the meta, creates the head, and puts the meta in it. Wrapping the meta in <head> of your own is unnecessary, and harmless where the source has a head already, because a second head is a parse error and dropped. Measured against x/net/html:
document the meta lands <!doctype html><html><head>… in the head <!doctype html><p>x</p> in the head - the parser creates one <!doctype html><html><body><p>x</p>… in the head <!doctype html><body><p>x</p> in the head <!doctype html><title>t</title><p>x</p> in the head, before the title <!doctype html><!-- c --><p>x</p> in the head <!doctype html>\n <p>x</p> in the head <!doctype html>text<p>x</p> in the *body*, where it is ignored <!doctype html> <p>x</p> in the body: nbsp is not whitespace just text in the body
So the rule is: the meta reaches the head if nothing but the doctype, comments and whitespace precedes the insertion point. The two failures are the same failure - text before the first element ends the head, and a document that is only text has no first element at all.
The non-breaking space is worth its own line. It looks like whitespace, it is not one of the five characters a parser skips before the head, and a template that indents its output with one by accident moves every meta tag into the body.
And that is detectable, in one pass, before the insertion ¶
Text arrives before the element that follows it, so a text handler knows whether the head is still open by the time the element handler runs. This program uses that: it counts non-whitespace text seen before the first element, inserts the meta anyway - a meta in the body is inert rather than harmful - and reports that a browser will ignore it. Reporting beats silence: a deploy id that is present in the source and invisible to the tooling that reads it is worse than one that is missing.
Source Files