Web scraping 黑客新闻：如何提取评论层次结构

我试图解析论坛news.ycombinator.com上的一条评论线索。然而，在查看html之后，似乎没有嵌套注释的层次结构。这将使解析变得非常困难。例如，下面是父注释及其子注释：

<!-- This part below draws the upvote/downvote images -->
<table border=0><tr><td><table border=0><tr><td><img src="http://ycombinator.com/images/s.gif" height=1 width=0></td><td valign=top><center><a id=up_4241971 href="vote?for=4241971&dir=up&whence=%69%74%65%6d%3f%69%64%3d%34%32%34%31%37%38%34"><img src="http://ycombinator.com/images/grayarrow.gif" border=0 vspace=3 hspace=2></a><span id=down_4241971></span></center></td><td class="default"><div style="margin-top:2px; margin-bottom:-10px; ">


<!-- This part below is user/time and permalink info for a parent comment -->
<span class="comhead"><a href="user?id=JshWright">JshWright</a> 7 hours ago  | <a href="item?id=4241971">link</a></span></div><br>


<!-- This part below is actual Comment -->
<span class="comment"><font color=#000000>I just got my Verizon Galaxy S3, and ordered the 20-pack of NFC tags offered by <a href="http://tagsfordroid.com" rel="nofollow">http://tagsfordroid.com</a><p>I think I know what my Dad felt like when he got his first label printer... Within days it seemed like every object in his office was labeled...<p>I've got a tag in my car to automatically send my wife a "Headed home" SMS, a tag on my night stand to toggle between 'night' (silent) and 'day' (loud) volume settings, a tag by my back door to launch CardioTrainer when I go out for a run (this one may have crossed the "I've run out of ideas" line...). I'm using the keychain tag to dial a response number for the fire department I'm a member of.</font></span><p><font size=1><u><a href="reply?id=4241971&whence=%69%74%65%6d%3f%69%64%3d%34%32%34%31%37%38%34">reply</a></u></font></td></tr></table></td></tr>


<!-- This part below is upvote/downvote arrow for child of parent -->
<tr><td><table border=0><tr><td><img src="http://ycombinator.com/images/s.gif" height=1 width=40></td><td valign=top><center><a id=up_4242025 href="vote?for=4242025&dir=up&whence=%69%74%65%6d%3f%69%64%3d%34%32%34%31%37%38%34"><img src="http://ycombinator.com/images/grayarrow.gif" border=0 vspace=3 hspace=2></a><span id=down_4242025></span></center></td><td class="default"><div style="margin-top:2px; margin-bottom:-10px; ">

<!-- This part has user/time/permalink for child comment -->
<span class="comhead"><a href="user?id=msbmsb">msbmsb</a> 7 hours ago  | <a href="item?id=4242025">link</a></span></div><br>

<!-- This part is the content of the  child comment -->
<span class="comment"><font color=#000000>I did the same thing. Tag next to the entry-way light switch for changing to an "at-home" profile, tag next to the bed for switching between night mode and morning mode, tag at work, keychain tag for switching between car mode and quiet mode.<p>And profile switching is just the basics. You can have a tag that connects guests' NFC-enabled phones to your wifi without having to hand out the password, for instance.<p>NFC task launcher + tasker is an amazing combination that opens up all kinds of possibilities.</font></span><p><font size=1><u><a href="reply?id=4242025&whence=%69%74%65%6d%3f%69%64%3d%34%32%34%31%37%38%34">reply</a></u></font></td></tr></table></td></tr><tr><td>

7小时前|

我刚买了Verizon Galaxy S3，订购了20包NFC标签，由提供。我想我知道我爸爸买第一台标签打印机时的感受。。。几天之内，他办公室里的每件物品都被贴上了标签……我的车上有一个标签，可以自动向我妻子发送“回家”短信；我的床头柜上有一个标签，可以在“夜间”（无声）和“白天”（大声）音量设置之间切换；我的后门有一个标签，可以在我出去跑步时启动心脏训练器（这一个可能已经越过了边界）“我没有主意了”行…。我正在使用钥匙链标签拨打我所属消防队的响应号码。
7小时前|

我也做了同样的事情。在入口通道灯开关旁边贴上标签，以便改为“在家”“配置文件，床边的标签用于在夜间模式和早晨模式之间切换，工作时的标签，钥匙链标签用于在汽车模式和安静模式之间切换。而配置文件切换只是基础。例如，您可以使用一个标签，将客人的支持NFC的手机连接到您的wifi，而无需提供密码。NFC task launcher+tasker是一个神奇的组合，它开启了各种可能性。

---

那么，hacker news如何存储评论的层次结构，以及我在抓取他们的数据时如何复制它？

在表中，缩进是通过图像标记完成的：

...<td><img src="http://ycombinator.com/images/s.gif" height=1 width=0></td>...
...<td><img src="http://ycombinator.com/images/s.gif" height=1 width=40></td>...

。。。。。。
......

---

大概你会阅读并解析这些。通过保持

width

值的内部堆栈，可以重建所表示的实际线程。

Wow！我错过了。谢谢。