---
 README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 323de86..5338dbf 100644
--- a/README.md
+++ b/README.md
@@ -4,8 +4,8 @@ aerc is an email client for your terminal.
 
 A short demonstration can be found on [https://aerc-mail.org/](https://aerc-mail.org/)
 
-Join the IRC channel: [#aerc on irc.freenode.net](http://webchat.freenode.net/?channels=aerc&uio=d4)
-for end-user support, and [#aerc-dev](http://webchat.freenode.net/?channels=aerc-dev&uio=d4)
[message trimmed]

Hi,

My two cents: I think adding a native type, and removing the stdlib
type, makes sense. I disagree that LLVM not having native support is a
compelling reason to avoid it. LLVM also lacks support for a lot of
other features Zig provides, which makes sense, because LLVM is a
*low-level IR*. Additionally, major decisions about the language
(whether a type should be builtin qualifies, IMO) should not be made
based on the limitations of the stage1 compiler, or the decisions of an
external project.

That said, having translate-c produce code using std.math.Complex would
also be an acceptable solution. However, it would be the first case of C
code that cannot automagically be translated to raw Zig code. This would

Hi, 

Just wanted to note that there's a difference between zig the language and
Zig the LLVM-based compiler, and give my two cents.

I'm currently working on a compiler[1] that supports both Zig and BTR as input
languages (though the BTR frontend has not yet been written, as I'm finishing
up ARM64 codegen first), is written in pure C99 (and is primarily tested on
plan9!), and with a focus on compiler simplicity and correctness.

Zig the language is more complex than BTR, but is arguably simpler than C in
many respects. Zig the compiler (both the LLVM-based stage1 and the self-hosted
compiler) are absolutely more complex than BTR, and small C compilers like cproc
or kencc, but that is an implementation concern; the language allows for simpler

+ new_path = file_rel_dir(ctx.path, ctx.token.ptr, ctx.token.len);
+ new_path_len = strlen(new_path);

Maybe make file_rel_dir take a pointer to len and store it if nonnull, to 
avoid recalculating an already known length.

+file_rel_dir(const char *const orig, const char *const new, size_t 
new_len)
+ out = malloc(last_slash_pos + strlen(new));

strlen(new) is illegal here, since new is not required to be null 
terminated. Furthermore, we have the length as new_len.

Note to self for tomorrow when i review this properly: children list in 
single processed queue is flat heaped list and should be freed as such.

I think it is intended to make indexing easier for Google.

---
 main.go | 131 ++++++++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 108 insertions(+), 23 deletions(-)

diff --git a/main.go b/main.go
index de770d6..c8af46b 100644
--- a/main.go
+++ b/main.go
@@ -142,7 +142,7 @@ func metricName(metric map[string]string) string {
 	return out + "{" + strings.Join(inner, ",") + "}"
 }
 
-func handleLabel(p *plot.Plot, l *plotter.Line, label string, tags map[string]string) {
+func handleLabel(p *plot.Plot, l plot.Thumbnailer, label string, tags map[string]string) {
[message trimmed]

---
 main.go | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)

diff --git a/main.go b/main.go
index 0f4b674..de770d6 100644
--- a/main.go
+++ b/main.go
@@ -47,7 +47,7 @@ type Datapoint struct {
 }
 
 type PromResult struct {
-	Metric string
+	Metric map[string]string
[message trimmed]

---
 main.go | 247 ++++++++++++++++++++++++++++++++------------------------
 1 file changed, 143 insertions(+), 104 deletions(-)

diff --git a/main.go b/main.go
index ab164e9..0f4b674 100644
--- a/main.go
+++ b/main.go
@@ -3,9 +3,11 @@ package main
 import (
 	"bytes"
 	"encoding/json"
+	"errors"
 	"fmt"
[message trimmed]

Rationale for this one:

The metric name was the only thing needed before I added label support, but the
label templating uses the raw tags, so it makes more sense now to preserve the
tags and render the name as needed instead of merging into into a string and
then splitting back into tags.